[{"data":1,"prerenderedAt":98},["ShallowReactive",2],{"$fu7saqfu966pj":3},{"href":4,"title":5,"description":6,"kind":7,"mark":7,"planned":8,"contributors":9,"provenance":7,"express":8,"bridge":8,"html":10,"headings":11},"\u002Fdocs\u002Ftools\u002Fworkspaces","Workspaces","A throwaway clone of one git repository for one run: set up by the repository itself, changed through the workspace tools, pushed to a branch, and deleted when the run ends.",null,false,[],"\u003Cp>Onboarding an existing project? Start with \u003Ca href=\"#what-the-repository-provides\">What the repository provides\u003C\u002Fa>, or give\nyour coding assistant the \u003Ca href=\"#llm-instruction-for-migrating-an-existing-workspace\">LLM instruction for migrating an existing workspace\u003C\u002Fa>.\nTo set up the whole team as well (token, MCP server, code host, team clone and schedules), follow\n\u003Ca href=\"\u002Fdocs\u002Fengineering-team\">An engineering team on your repository\u003C\u002Fa>.\u003C\u002Fp>\n\u003Ch2 id=\"what-a-workspace-is\">What a workspace is\u003C\u002Fh2>\n\u003Cp>A workspace is a clone of one git repository, made for one run and deleted when that run ends. A member reads,\nchanges, tests and commits code in it, and pushes the result to a branch. Nothing in it outlives the run except\nwhat was pushed.\u003C\u002Fp>\n\u003Cp>A run gets a workspace only when its task declares workspace tools: \u003Ccode>workspace.read_file\u003C\u002Fcode>, \u003Ccode>workspace.edit\u003C\u002Fcode>,\n\u003Ccode>workspace.run_command\u003C\u002Fcode>, \u003Ccode>workspace.git_push\u003C\u002Fcode> and the rest, each listed by name like any other tool. A task\nwithout them never clones anything. What each tool takes and returns is in the\n\u003Ca href=\"\u002Fdocs\u002Ftools\u002Fbuilt-in\u002Fworkspace\">workspace tools reference\u003C\u002Fa>.\u003C\u002Fp>\n\u003Cp>Everything a workspace does is plain git and the repository's own tooling: it clones, branches, commits, merges and\npushes through the repository's remote, and sets up and checks the code with the commands the repository declares\n(see \u003Ca href=\"#what-the-repository-provides\">What the repository provides\u003C\u002Fa>). None of that depends on where the repository\nis hosted: the remote's address says which host it is, and the credential for it comes from the code host tool you\nbind. The task that reads an issue or pull request supplies its branch name. See\n\u003Ca href=\"#where-the-repository-lives\">Where the repository lives\u003C\u002Fa> and \u003Ca href=\"#the-branch\">The branch\u003C\u002Fa>.\u003C\u002Fp>\n\u003Cp>The workspace tools only touch the clone. Issues, pull requests, reviews and merging happen on your code host,\nthrough the host's own MCP server, bound like any other tool: \u003Ca href=\"\u002Fdocs\u002Ftools\u002Fgithub\">GitHub\u003C\u002Fa> has a setup guide, and\nany other host's server can be bound from \u003Ca href=\"\u002Fdocs\u002Ftools\">Tools\u003C\u002Fa>.\u003C\u002Fp>\n\u003Ch2 id=\"where-the-repository-lives\">Where the repository lives\u003C\u002Fh2>\n\u003Cp>A run names its repository by its git remote address, on any host that serves git: GitLab, Bitbucket, Azure\nRepos, a self-hosted server, or GitHub.\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-json\">{ &quot;repo&quot;: &quot;https:\u002F\u002Fgit.example.com\u002Facme\u002Fwebsite.git&quot; }\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Nothing about the host is set in the OS. The workspace reads what it needs from the remote itself: the host from\nthe address, and the default branch and the branches from git.\u003C\u002Fp>\n\u003Cp>The repository is cloned the first time the run needs it, with its last 50 commits.\u003C\u002Fp>\n\u003Ch3 id=\"the-credential-comes-from-the-code-host-tool-you-bind\">The credential comes from the code host tool you bind\u003C\u002Fh3>\n\u003Cp>The clone and every push use the code host tool your enterprise or team binds on \u003Ca href=\"\u002Fdocs\u002Ftools\">Tools\u003C\u002Fa>. A workspace\nreads these custom props off the binding: key → value settings added behind \u003Cstrong>Custom props\u003C\u002Fstrong> on its form.\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Key\u003C\u002Fth>\n\u003Cth>What it is\u003C\u002Fth>\n\u003Cth>GitHub's\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>git.host\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The host this binding's token is valid for\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>github.com\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>git.username\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The username git sends alongside the token\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>x-access-token\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>git.repoTemplate\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>A clone address with \u003Ccode>{repo}\u003C\u002Fcode> in it, applied to a short \u003Ccode>owner\u002Fname\u003C\u002Fcode> input\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>https:\u002F\u002Fgithub.com\u002F{repo}.git\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>git.repo\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The repository the binding's layer works in, for a run whose input names none\u003C\u002Ftd>\n\u003Ctd>Usually left out\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>The workspace takes the binding, among those the run reaches (task, role, member, team, enterprise, in that\norder), whose \u003Ccode>git.host\u003C\u002Fcode> matches the remote's. Each tool counts once: the nearest binding of it the run reaches, so\na member's own binding without the props hides the enterprise's. Its token is given to git scoped to that host, and only over\n\u003Ccode>https:\u002F\u002F\u003C\u002Fcode>, so a prompt from any other host is answered with nothing. A remote on a host no binding names is cloned\nwith no credential: a public repository still clones, and nothing pushes. The \u003Ca href=\"\u002Fdocs\u002Ftools\u002Fgithub\">GitHub\u003C\u002Fa> guide\nsays what to set for GitHub.\u003C\u002Fp>\n\u003Cp>A binding may also name the repository its layer works in (\u003Ccode>git.repo\u003C\u002Fcode>), and then a run needs no \u003Ccode>repo\u003C\u002Fcode> input at\nall; a run that gives one overrides it. Because a binding can carry \u003Ccode>git.repoTemplate\u003C\u002Fcode>, a short input keeps working:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-json\">{ &quot;repo&quot;: &quot;acme\u002Fwebsite&quot; }\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Ch3 id=\"credentials-for-the-repositorys-own-commands\">Credentials for the repository's own commands\u003C\u002Fh3>\n\u003Cp>A task that deploys, applies infrastructure or publishes runs the repository's own commands, and those commands need\nthe enterprise's keys: a hosting token, a cloud credential, a state store's access key. The task version names what\neach command reads and which bound tool it comes from, in its workspace:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-json\">{\n    &quot;credentials&quot;: [\n        { &quot;tool&quot;: &quot;railway&quot;, &quot;as&quot;: &quot;RAILWAY_TOKEN&quot; },\n        { &quot;tool&quot;: &quot;terraform-state&quot;, &quot;as&quot;: &quot;AWS_ACCESS_KEY_ID&quot;, &quot;prop&quot;: &quot;accessKeyId&quot; },\n        { &quot;tool&quot;: &quot;terraform-state&quot;, &quot;as&quot;: &quot;AWS_SECRET_ACCESS_KEY&quot; }\n    ]\n}\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>\u003Ccode>tool\u003C\u002Fcode> is the prefix of a tool bound on \u003Ca href=\"\u002Fdocs\u002Ftools\">Tools\u003C\u002Fa>; \u003Ccode>as\u003C\u002Fcode> is the environment variable the command reads;\n\u003Ccode>prop\u003C\u002Fcode> takes one of the binding's custom props instead of its secret. The value comes from the nearest binding of\nthat tool the run reaches, of any kind: a \u003Cstrong>Secret\u003C\u002Fstrong> stored on Tools for exactly this (choose the Secret kind, give it\nthe prefix, the value and any custom props), a web credential, or a server's token.\u003C\u002Fp>\n\u003Cp>The value goes into the environment of \u003Ccode>workspace.run_command\u003C\u002Fcode> and nowhere else. The model never reads it, and\nwhatever a command prints has every value masked (\u003Ccode>[credential removed]\u003C\u002Fcode>) before the model or the run's record sees\nit. A name the harness owns (\u003Ccode>PATH\u003C\u002Fcode>, \u003Ccode>HOME\u003C\u002Fcode>, \u003Ccode>TMPDIR\u003C\u002Fcode>, \u003Ccode>TMP\u003C\u002Fcode>, \u003Ccode>TEMP\u003C\u002Fcode>, \u003Ccode>GIT_*\u003C\u002Fcode>, \u003Ccode>OS_*\u003C\u002Fcode>) is refused when the version is saved.\u003C\u002Fp>\n\u003Cp>A run whose version asks for a credential no binding holds fails \u003Ccode>workspace_credential_missing\u003C\u002Fcode> before it calls\nanything, and is listed under Blockers as a credentials block: save the binding on Tools and it resumes.\u003C\u002Fp>\n\u003Cp>A run with no repository — neither in its input nor on a binding it reaches — fails \u003Ccode>workspace_repo_missing\u003C\u002Fcode>\nbefore it calls anything. A run that names a short \u003Ccode>owner\u002Fname\u003C\u002Fcode> no binding's \u003Ccode>git.repoTemplate\u003C\u002Fcode> expands fails\n\u003Ccode>workspace_repo_no_clone_address\u003C\u002Fcode> when it opens its workspace: the OS never guesses a host for it. Tools says when\ntasks work in a workspace and no binding has \u003Ccode>git.*\u003C\u002Fcode> props, and flags a binding without them when another binding\nof the same tool has them.\u003C\u002Fp>\n\u003Ch2 id=\"the-branch\">The branch\u003C\u002Fh2>\n\u003Cp>A run works on one branch, chosen in this order:\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>The run works on\u003C\u002Fth>\n\u003Cth>When\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>Its \u003Ccode>branch\u003C\u002Fcode> input\u003C\u002Ftd>\n\u003Ctd>The run was started with one, or the task before it handed one on.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>The task's branch template\u003C\u002Ftd>\n\u003Ctd>The task's workspace sets \u003Ccode>branch\u003C\u002Fcode> and the input fills every \u003Ccode>{key}\u003C\u002Fcode>: \u003Ccode>fix\u002F{issue}\u003C\u002Fcode> with \u003Ccode>{ &quot;issue&quot;: 42 }\u003C\u002Fcode> is \u003Ccode>fix\u002F42\u003C\u002Fcode>.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os\u002F&lt;task slug&gt;-&lt;first 8 characters of the run id&gt;\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Neither. A new branch.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>The default branch's tip, under that same new name\u003C\u002Ftd>\n\u003Ctd>The task's workspace sets \u003Ccode>onDefaultBranch: true\u003C\u002Fcode>. No input or template is read: the run works on the default branch as it is, nothing is merged into it, and it cannot push. For a task that reads or verifies the default branch itself.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>The workspace does not look up an issue or pull request to choose a branch. The task that reads it reports a\n\u003Ccode>branch\u003C\u002Fcode> in its result, and the following tasks, retries and resumes inherit it while working on the same\nrepository and issue or pull request. Renaming an issue later does not rename the branch. If a template needs\nan input that is missing (including \u003Ccode>issue_slug\u003C\u002Fcode>), the run uses the fallback branch above. A run that started\nwithout a branch also passes the branch it was given to its successors.\u003C\u002Fp>\n\u003Cp>A branch that exists on the remote is checked out as it is there. One that does not is created from the default\nbranch: the branch the clone checks out, which is the repository's own default.\u003C\u002Fp>\n\u003Ch3 id=\"kept-up-with-the-default-branch\">Kept up with the default branch\u003C\u002Fh3>\n\u003Cp>When the run's branch already exists and is behind the default branch, the default branch is merged into it and\npushed before setup runs and before the member does anything, so an open pull request never falls more than one\nrun behind. It is an ordinary merge and an ordinary push. If the merge conflicts, or the remote refuses the push, it\nis undone and the branch is left exactly as the remote has it, for the run to deal with.\u003C\u002Fp>\n\u003Ch2 id=\"what-the-repository-provides\">What the repository provides\u003C\u002Fh2>\n\u003Cp>How the code is set up and checked belongs to the repository, not to the task. Tasks are shared between\nenterprises through the catalog and pointed at other repositories, so a task's skill never names one repository's\nscripts.\u003C\u002Fp>\n\u003Ch3 id=\"setup\">Setup\u003C\u002Fh3>\n\u003Cp>After cloning, the workspace runs the repository's setup. With no setup file, it works out what to run, and runs\nboth lines when both apply:\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>The repository has\u003C\u002Fth>\n\u003Cth>Setup runs\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise.toml\u003C\u002Fcode> or \u003Ccode>.tool-versions\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise install\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>package.json\u003C\u002Fcode> with a \u003Ccode>packageManager\u003C\u002Fcode> of npm, and no \u003Ccode>pnpm-lock.yaml\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>npm ci\u003C\u002Fcode>, or \u003Ccode>npm install\u003C\u002Fcode> without \u003Ccode>package-lock.json\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>Any other \u003Ccode>package.json\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>pnpm install --frozen-lockfile\u003C\u002Fcode>, or \u003Ccode>pnpm install\u003C\u002Fcode> without \u003Ccode>pnpm-lock.yaml\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>To choose for yourself, add \u003Ccode>.os\u002Fworkspace.yml\u003C\u002Fcode> (or \u003Ccode>.os\u002Fworkspace.yaml\u003C\u002Fcode>, or \u003Ccode>.os\u002Fworkspace.json\u003C\u002Fcode>) at the root of\nthe repository:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-yaml\"># Leave setup out to use the list above. `setup: []` skips setup altogether.\nsetup:\n    - [pnpm, install, --frozen-lockfile]\n    - [pnpm, run, generate]\nteardown:\n    - [node, scripts\u002Fclean-up.mjs]\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cul>\n\u003Cli>Each step is one line, \u003Ccode>- [binary, argument, …]\u003C\u002Fcode>: no shell, and no other YAML form. In this form an argument\ncannot contain a comma. Use the JSON form for arguments needing commas or more complex quoting.\u003C\u002Fli>\n\u003Cli>The binary is one of \u003Ccode>git\u003C\u002Fcode>, \u003Ccode>node\u003C\u002Fcode>, \u003Ccode>npm\u003C\u002Fcode>, \u003Ccode>pnpm\u003C\u002Fcode> and \u003Ccode>mise\u003C\u002Fcode>. Anything else a repository needs (another\nlanguage, a particular version of Node) comes through mise: declare it in \u003Ccode>mise.toml\u003C\u002Fcode>, and \u003Ccode>mise install\u003C\u002Fcode>\ninstalls it.\u003C\u002Fli>\n\u003Cli>\u003Ccode>setup\u003C\u002Fcode> runs in order after every clone. If a step exits non-zero, the clone is thrown away and the workspace\ncall fails with \u003Ccode>workspace_setup_failed\u003C\u002Fcode>, naming the binary and the start of its output. The member is handed\nthe error, and its next workspace call tries again.\u003C\u002Fli>\n\u003Cli>\u003Ccode>teardown\u003C\u002Fcode> runs when the workspace is deleted. A step that fails there is ignored.\u003C\u002Fli>\n\u003Cli>The JSON form is the same two lists: \u003Ccode>{ &quot;setup&quot;: [[&quot;pnpm&quot;, &quot;install&quot;]], &quot;teardown&quot;: [] }\u003C\u002Fcode>.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>An explicit \u003Ccode>setup\u003C\u002Fcode> \u003Cstrong>replaces all auto-detection\u003C\u002Fstrong>. Include \u003Ccode>mise install\u003C\u002Fcode> yourself if your setup needs tools\ndeclared in mise. There is no automatic \u003Ccode>os:setup\u003C\u002Fcode> or task-level startup hook: if you define an \u003Ccode>os:setup\u003C\u002Fcode> mise\ntask, call it explicitly from this file. Steps run from the repository root. Put shell logic and subdirectory\ncommands in repository scripts or mise tasks, then invoke those through an allowed binary.\nKeep only one manifest: if several exist, \u003Ccode>.yml\u003C\u002Fcode> wins over \u003Ccode>.yaml\u003C\u002Fcode>, then \u003Ccode>.json\u003C\u002Fcode>.\u003C\u002Fp>\n\u003Cp>For Yarn, Bun or a non-Node stack, provide explicit setup; auto-detection does not select those package managers.\nUse locked, reproducible installs, generate anything a clean checkout needs, and install the repository's git hooks.\nThe environment is non-interactive, with \u003Ccode>CI=true\u003C\u002Fcode>, \u003Ccode>NODE_ENV=test\u003C\u002Fcode> and a fresh home directory. Do not depend on\nyour laptop's shell profile, global packages, local \u003Ccode>.env\u003C\u002Fcode> files or an existing database. Document any private\npackage access, test services or other prerequisites; binding a code host does not configure those for you.\u003C\u002Fp>\n\u003Cp>After that setup, if the clone depends on \u003Ccode>@playwright\u002Ftest\u003C\u002Fcode> or \u003Ccode>playwright\u003C\u002Fcode>, the OS installs Chromium for the\ninstalled version into a shared cache (\u003Ccode>&lt;runnerRoot&gt;\u002F.playwright-browsers\u002F&lt;version&gt;\u003C\u002Fcode>) and sets\n\u003Ccode>PLAYWRIGHT_BROWSERS_PATH\u003C\u002Fcode> on every workspace command. A cache hit is reused; a failed install is a warning on\n\u003Ccode>run.workspace_ready\u003C\u002Fcode> and does not fail setup. The install does not run apt (\u003Ccode>--with-deps\u003C\u002Fcode>): the run-host image\nmust already include Playwright's Chromium system libraries (the production worker image does). Set\n\u003Ccode>playwright: false\u003C\u002Fcode> in \u003Ccode>.os\u002Fworkspace.yml\u003C\u002Fcode> to skip this, even when the package is present. Workspace commands still\nget \u003Ccode>PLAYWRIGHT_SKIP_BROWSER_DOWNLOAD=1\u003C\u002Fcode> then, so a repository that opts out installs its browser in its own setup\n(for example \u003Ccode>npm exec -- playwright install chromium\u003C\u002Fcode>).\u003C\u002Fp>\n\u003Ch3 id=\"services-a-database-for-each-run\">Services: a database for each run\u003C\u002Fh3>\n\u003Cp>A repository whose commands need a real database, to prove its migrations for example, can ask for one in the same\nfile:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-yaml\">services: [mysql]\nsetup:\n    - [pnpm, install, --frozen-lockfile]\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>\u003Ccode>services\u003C\u002Fcode> is a list: \u003Ccode>services: [mysql]\u003C\u002Fcode> on one line, \u003Ccode>services:\u003C\u002Fcode> with one \u003Ccode>- mysql\u003C\u002Fcode> per line below it, or\n\u003Ccode>&quot;services&quot;: [&quot;mysql&quot;]\u003C\u002Fcode> in the JSON form. Leave it out and the run gets none. The one service today is \u003Ccode>mysql\u003C\u002Fcode>. Any\nother name fails the setup with \u003Ccode>workspace_service_unknown:&lt;name&gt;\u003C\u002Fcode>. The OS never skips a service it does not know.\u003C\u002Fp>\n\u003Cp>For \u003Ccode>mysql\u003C\u002Fcode>, each run gets an empty MySQL database of its own and a user with every privilege on that database and\non no other. Both are made before any of the repository's commands run, so the \u003Ccode>setup\u003C\u002Fcode> steps can use them as well as\nthe commands a task runs. Every command has exactly these five variables:\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Variable\u003C\u002Fth>\n\u003Cth>What it holds\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>WORKSPACE_MYSQL_HOST\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The database server's host\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>WORKSPACE_MYSQL_PORT\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Its port\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>WORKSPACE_MYSQL_USER\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The run's own user\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>WORKSPACE_MYSQL_PASSWORD\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>That user's password, new for each run\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>WORKSPACE_MYSQL_DATABASE\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The run's own database\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>Read them in your own scripts, for example to build a connection URL for your migration tool. The password is\nremoved from anything a command prints before a member or the run's record sees it. The database is never the OS's\nown, and a run cannot reach any other run's.\u003C\u002Fp>\n\u003Cp>The database and the user are destroyed with the workspace when the run ends, whether it succeeded or failed, after\nthe \u003Ccode>teardown\u003C\u002Fcode> steps. Anything a run leaves behind is dropped once it is six hours old. Nothing in it outlasts the run,\nso a run starts from an empty database every time.\u003C\u002Fp>\n\u003Cp>If the deployment has no database server for workspaces, or a run's workspace cannot reach it, the setup fails with\n\u003Ccode>workspace_service_unavailable:mysql\u003C\u002Fcode>, and says why, before any of the repository's commands run. The run is blocked\nand shown on Blockers. It never runs without the database the repository asked for.\u003C\u002Fp>\n\u003Ch3 id=\"wiring-mise-toml\">Wiring \u003Ccode>mise.toml\u003C\u002Fcode>\u003C\u002Fh3>\n\u003Cp>Keep the project's existing language, package manager and test framework. Mise supplies a common command name\nfor each operation; the implementation stays in your repository. Add to an existing \u003Ccode>mise.toml\u003C\u002Fcode> rather than\nreplacing it. This tasks-only Node example delegates to scripts you must implement using the contracts below:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-toml\">[tasks.&quot;os:red&quot;]\ndescription = &quot;Run changed tests under tests\u002F with the project's test framework; print a short result (a line per test file and per failing test, full output to a file) and preserve the RED exit codes.&quot;\nrun = &quot;node scripts\u002Fos-red.mjs&quot;\n\n[tasks.&quot;os:check&quot;]\ndescription = &quot;Run lint, typecheck, tests and prerequisite builds; accept an optional issue number.&quot;\nrun = &quot;node scripts\u002Fos-check.mjs&quot;\n\n[tasks.&quot;os:coverage&quot;]\ndescription = &quot;Measure coverage of changed source lines against this project's thresholds.&quot;\nrun = &quot;node scripts\u002Fos-coverage.mjs&quot;\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Replace the descriptions, paths and commands with your project's real ones. Trailing arguments must reach the\nscripts: \u003Ccode>os:red\u003C\u002Fcode> accepts an issue number; \u003Ccode>os:check\u003C\u002Fcode> and \u003Ccode>os:coverage\u003C\u002Fcode> accept an optional one. Wrappers must\npreserve exit codes. A task that just prints success does not implement a hook.\u003C\u002Fp>\n\u003Cp>If you need pinned runtimes, add a \u003Ccode>[tools]\u003C\u002Fcode> table using versions compatible with the project. Run tools through\n\u003Ccode>mise run\u003C\u002Fcode> or \u003Ccode>mise exec\u003C\u002Fcode> when they need mise's environment; installation alone does not activate a later bare\ncommand. Check configuration trust in a fresh, non-interactive home directory. A reviewed config that requires\ntrust can be trusted explicitly in setup with \u003Ccode>[mise, trust, mise.toml]\u003C\u002Fcode> before installation. Keep complex shell\nlogic in scripts so template delimiters in inline commands do not unexpectedly trigger mise's trust checks.\nSee mise's \u003Ca href=\"https:\u002F\u002Fmise.jdx.dev\u002Ftasks\u002Ftask-configuration.html\">task configuration\u003C\u002Fa> and\n\u003Ca href=\"https:\u002F\u002Fmise.jdx.dev\u002Fcli\u002Ftrust.html\">trust command\u003C\u002Fa>.\u003C\u002Fp>\n\u003Ch3 id=\"git-hooks\">Git hooks\u003C\u002Fh3>\n\u003Cp>The repository's own git hooks run on the commits and pushes a member makes, once setup has installed them (a\n\u003Ccode>prepare\u003C\u002Fcode> script that installs hooks runs as part of \u003Ccode>pnpm install\u003C\u002Fcode> or \u003Ccode>npm ci\u003C\u002Fcode>). The workspace tools never skip\nthem, and refuse the usual ways around them. A hook that fails refuses the commit or the push, and its output goes\nback to the member to fix.\u003C\u002Fp>\n\u003Cp>The engineering workflow commits and pushes deliberately failing tests before implementing them. Keep static\nchecks (formatting, lint and typecheck) working at that stage, while allowing genuine RED assertion failures to\nreach the branch. If your hooks require every test to pass before any commit or push, adapt that policy explicitly\nfor the RED stage. Keep full tests and coverage before merge, in the verification below. Never solve this with\n\u003Ccode>--no-verify\u003C\u002Fcode>, \u003Ccode>HUSKY=0\u003C\u002Fcode>, skipped tests or a blanket hook bypass.\u003C\u002Fp>\n\u003Ch3 id=\"verification-before-merge\">Verification before merge\u003C\u002Fh3>\n\u003Cp>The engineering tasks verify a head in the workspace and record it on the pull request. The task that ran the\nrepository's check (\u003Ccode>os:check\u003C\u002Fcode>, part by part, and \u003Ccode>os:coverage\u003C\u002Fcode>) posts one note per head it checked, headed\n\u003Ccode>## Verification: &lt;commit&gt;\u003C\u002Fcode>, naming each part and its exit, the coverage verdict, and the run that did it. The merge\ntask reads the newest note for the head it would merge before it asks for approval, and names the verified commit in\nwhat the approver reads. A head nobody verified, such as a push a person made or the head after an update from the\ndefault branch, is verified by a task of its own first. The OS runs no CI and reads no check run; a repository with\nits own CI may have its tasks read that too. \u003Ccode>workspace.git_push\u003C\u002Fcode> answers the branch and the commit it pushed, so the\nnote names exactly what went up.\u003C\u002Fp>\n\u003Ch3 id=\"commands-for-the-engineering-tasks\">Commands for the engineering tasks\u003C\u002Fh3>\n\u003Cp>The engineering tasks in the catalog (writing an issue's tests, implementing it, validating it) never run a\nrepository's scripts by name. They call three \u003Ca href=\"https:\u002F\u002Fmise.jdx.dev\">mise\u003C\u002Fa> tasks, and the repository's\n\u003Ccode>mise.toml\u003C\u002Fcode> says what each one does:\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Command\u003C\u002Fth>\n\u003Cth>Called as\u003C\u002Fth>\n\u003Cth>What it has to do\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:red\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:red &lt;issue&gt;\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Run the issue's tests: the test files the branch adds or changes against the default branch, including uncommitted tests. Print a short result a task can read whole, however many packages the tests touch: one line per test file with its path and its failed and passed counts, one line per failing test with its package, its name and the first line of its failure message, no passing tests, stack traces or colour codes, and the exit code last. Keep it under 10,000 characters; past that, count the failing tests left out on one line. Write the test runners' full output to a file, and name that file in the result. Exit 0 when every test passes, 1 for assertion failures, 3 for no suite (\u003Ccode>red_suite_missing\u003C\u002Fcode>), 4 for typecheck failures (\u003Ccode>red_suite_typecheck\u003C\u002Fcode>), and 5 for skipped, focused or todo tests (\u003Ccode>red_suite_skips\u003C\u002Fcode>). Its mise \u003Ccode>description\u003C\u002Fcode> says where tests live and which framework runs them: the task that writes tests reads it.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:check\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:check [issue]\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Everything to run once before committing: lint, typecheck, tests, and whatever build they need first. Exit 0 when it all passes.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:coverage\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:coverage [issue]\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Measure test coverage of the source lines the branch adds or changes, against the repository's own thresholds. Exit 0 when they are met or no measurable source changed; otherwise exit non-zero and list uncovered changed lines per file.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:test:summary\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:test:summary\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Optional. Print the test evidence the last \u003Ccode>os:red\u003C\u002Fcode>, \u003Ccode>os:check\u003C\u002Fcode> and \u003Ccode>os:coverage\u003C\u002Fcode> runs left in the clone, as a markdown table for a pull request: one row per phase and command, with the result, the passed, failed, skipped and todo counts from the test runners' own reports, the exit code, the duration and the commit tested. Read-only: it runs no test. A command that left no report reads \u003Ccode>no_test_results\u003C\u002Fcode>, never zero passing.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>A repository that has \u003Ccode>os:test:summary\u003C\u002Fcode> keeps that evidence by running the three commands through a test-report\nwrapper, such as \u003Ccode>os-test-report\u003C\u002Fcode> from \u003Ccode>@juicyllama\u002Frepo\u003C\u002Fcode> 1.3.0:\n\u003Ccode>os-test-report run --phase red|check|coverage --label 'mise run os:&lt;name&gt;' -- &lt;the command&gt;\u003C\u002Fcode>. The wrapper keeps the\ncommand's exit status, and leaves \u003Ccode>.os\u002Ftest-results\u002F&lt;run&gt;\u002F\u003C\u002Fcode> (ignore it in git) with the full \u003Ccode>output.log\u003C\u002Fcode>, the\nrunners' native reports and the counts. It sets \u003Ccode>OS_TEST_REPORT_DIR\u003C\u002Fcode> for the command, so the command writes each\nsuite's report there: the runners' own JSON reports, and for a merge-base rerun a \u003Ccode>baseline\u002F\u003C\u002Fcode> directory, which is not\ncounted. A check that replays a cached result writes no report, so read nothing from a build cache for the test\nruns. \u003Ccode>os:red\u003C\u002Fcode>'s exits 3, 4 and 5 read \u003Ccode>red_suite_missing\u003C\u002Fcode>, \u003Ccode>red_suite_typecheck\u003C\u002Fcode> and \u003Ccode>red_suite_skips\u003C\u002Fcode>, 6\n\u003Ccode>pre_existing_failure\u003C\u002Fcode> and 7 \u003Ccode>environment_failure\u003C\u002Fcode>. The tasks that post a pull request's test evidence read\n\u003Ccode>os:test:summary\u003C\u002Fcode> when a repository has it, and say native counts are unavailable when it does not.\u003C\u002Fp>\n\u003Cp>Use the remote's actual default branch, not a hard-coded \u003Ccode>main\u003C\u002Fcode>. Comparisons need its merge base with the working\nbranch. A shallow clone may need a fetch or deeper history before that can be calculated; a missing base is an\nerror, never an empty diff or a passing check. The issue number identifies the work, not a required test filename.\nTests should use the repository's normal locations and naming conventions. A RED test must import and typecheck\nsuccessfully, then fail on an assertion; missing modules, syntax errors and crashed runners are not valid RED proof.\u003C\u002Fp>\n\u003Cp>\u003Ccode>os:check\u003C\u002Fcode> and \u003Ccode>os:coverage\u003C\u002Fcode> may also support these diagnostic exits:\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Exit\u003C\u002Fth>\n\u003Cth>When it is valid\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>6\u003C\u002Ftd>\n\u003Ctd>Every failure is a test proved to fail on the branch's merge base with the default branch too. Print \u003Ccode>pre_existing_failure: &lt;file&gt; › &lt;test&gt; (fails on &lt;sha&gt;)\u003C\u002Fcode> for each. Changed test files, broken suites, lint and typecheck failures do not qualify.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>7\u003C\u002Ftd>\n\u003Ctd>The test runner itself crashed, a retry also crashed, and there is no failure attributable to the branch. Print \u003Ccode>runner_crashed: &lt;task&gt;: &lt;reason&gt;\u003C\u002Fcode>. An exception in application code is not a runner crash. Also valid when no test runner started because the host had no room for one within a bounded wait, and nothing failed: print a \u003Ccode>host_short:\u003C\u002Fcode> line naming the free and maximum slots.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>Implement these only when the scripts can establish that evidence. Otherwise return an ordinary failure. Never\nrelabel a failure just to make the workflow continue.\u003C\u002Fp>\n\u003Cp>A repository with no \u003Ccode>os:red\u003C\u002Fcode> stops those tasks with a message saying it needs one. Any task can run whatever else\nthe repository's \u003Ccode>mise.toml\u003C\u002Fcode> defines, through \u003Ccode>workspace.run_command\u003C\u002Fcode>.\u003C\u002Fp>\n\u003Ch3 id=\"a-check-has-to-fit-one-command\">A check has to fit one command\u003C\u002Fh3>\n\u003Cp>Every command a member runs has \u003Ca href=\"#limits\">limits\u003C\u002Fa>: ten minutes, and 32,000 characters of output. A check that runs\npast the first is stopped before it can say what failed, and one that prints past the second is cut short. So:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Print what failed, not everything.\u003C\u002Fstrong> Keep the full output in a file in the clone and print where it is. A member\ncan read the file when it needs to.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Split a check that cannot finish in ten minutes.\u003C\u002Fstrong> \u003Ccode>mise run os:check --parts\u003C\u002Fcode> prints the name of each part, one\nper line, and \u003Ccode>mise run os:check --part &lt;name&gt;\u003C\u002Fcode> runs that part over the whole repository, with the same exit codes.\nA task that checks a whole branch asks for the parts and runs each in turn. A task that checks one issue asks with\nthe issue, \u003Ccode>mise run os:check --parts &lt;issue&gt;\u003C\u002Fcode>: accept the number there, and print either every part or only the\nparts that issue's branch needs. A repository that does not answer \u003Ccode>--parts\u003C\u002Fcode> is checked with one\n\u003Ccode>mise run os:check\u003C\u002Fcode>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Run each test once when coverage runs too.\u003C\u002Fstrong> A task that runs \u003Ccode>os:coverage\u003C\u002Fcode> after \u003Ccode>os:check\u003C\u002Fcode> asks\n\u003Ccode>mise run os:check --parts --covered-by-coverage &lt;issue&gt;\u003C\u002Fcode>. A repository whose coverage measurement runs some of\nthe same tests, and fails on a failing test with the same exit codes, leaves those test parts out of the list. A\nrepository that does not know the flag refuses it, and the task asks again without it.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Split a coverage measurement the same way.\u003C\u002Fstrong> \u003Ccode>mise run os:coverage --parts\u003C\u002Fcode> prints the parts for the branch as it\nis, one per line in the order to run them, and \u003Ccode>mise run os:coverage --part &lt;name&gt;\u003C\u002Fcode> runs one. Every part but the\nlast only measures. The last prints the verdict, with the exit codes of the whole command, and must refuse to judge\na part that has not run or that measured the branch before a later change. A repository that does not answer\n\u003Ccode>--parts\u003C\u002Fcode> is measured with one \u003Ccode>mise run os:coverage\u003C\u002Fcode>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Measure with the tests that can reach the change.\u003C\u002Fstrong> A measurement of the lines a branch changed learns nothing\nfrom a test that never loads a changed file. A repository can run only the tests that can, and should when its\nwhole suite takes longer than a part. The tests it leaves out still run in \u003Ccode>os:check\u003C\u002Fcode>, so it must not ask\n\u003Ccode>--covered-by-coverage\u003C\u002Fcode> to leave their parts out.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Keep every part well inside the limit.\u003C\u002Fstrong> A part is as long as the longest thing in it that cannot be divided. When\none test file alone takes most of ten minutes, adding parts does not help: run that file's tests across several\nparts, each test in exactly one. The last part must still fail on a test that failed in any part.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch3 id=\"a-command-plans-for-four-cpus\">A command plans for four CPUs\u003C\u002Fh3>\n\u003Cp>A workspace shares its machine with other runs. Every command is held to a CPU budget of 4, so a build or a test\nrunner that starts one worker per CPU starts four, not one for every core of the host.\u003C\u002Fp>\n\u003Cp>Most tools see the budget without being told: \u003Ccode>nproc\u003C\u002Fcode>, Node's \u003Ccode>os.availableParallelism()\u003C\u002Fcode>, Go, Rust and the JVM count\nthe CPUs a command may run on. Node's \u003Ccode>os.cpus()\u003C\u002Fcode> and \u003Ccode>os.availableParallelism()\u003C\u002Fcode> do too: every spawn command is\nstarted with a preload in \u003Ccode>NODE_OPTIONS\u003C\u002Fcode> that lists the CPUs it may run on, or \u003Ccode>OS_CPU_BUDGET\u003C\u002Fcode> when that is smaller,\nahead of any \u003Ccode>NODE_OPTIONS\u003C\u002Fcode> your command sets. Some tools count the\nmachine's cores anyway, such as Python's \u003Ccode>os.cpu_count()\u003C\u002Fcode>, and a tool built on those still sizes itself from the host.\nFor those, every command's environment carries \u003Ccode>OS_CPU_BUDGET\u003C\u002Fcode>. Read it where the tool takes its worker count, in its\nconfig or in your \u003Ccode>mise run os:*\u003C\u002Fcode> task:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-js\">const workers = Number(process.env.OS_CPU_BUDGET) || undefined\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>A command that starts far more processes and threads than that can run the machine out of them. It then fails with\n\u003Ccode>EAGAIN\u003C\u002Fcode> or \u003Ccode>Resource temporarily unavailable\u003C\u002Fcode> on code that builds anywhere else.\u003C\u002Fp>\n\u003Cp>The machine's process and thread limit is shared by every run on it. A machine near its limit starts no new run until\nit has room again, and another machine takes the run. A run with a command the machine had no room to start is\nnot blocked: its work is pushed to its branch, and it goes back to the queue by itself, to carry on where it stopped\non a machine with room. Its history says \u003Ccode>workspace_resource_exhausted:EAGAIN\u003C\u002Fcode>, with how many processes and threads\nthe machine held of how many it allows: \u003Ccode>workspace_resource_exhausted:EAGAIN pids.current=998 pids.max=1000\u003C\u002Fcode>. A run\nthat meets it five times in a row is one whose own commands need more than a machine has, and it waits on Blockers.\nThe same immediate path is taken when a command starts and then dies because the machine had no room: a test runner\ncrash, a browser that could not launch or crashed under load, or a heap limit. A test part that never starts because\nthe host is too busy (\u003Ccode>host_short:\u003C\u002Fcode>, exit 7) is classed \u003Ccode>workspace_resource_exhausted:HOSTSHORT\u003C\u002Fcode> and requeues later,\non a backoff of 2, 5, 15, 30 and then 60 minutes, instead: retrying at once would land on the same busy machine. It\nwaits on Blockers, and is escalated, only after all five of those retries are spent. A command that succeeded, or\none whose output names a failing test, always comes back to the run as it was, whatever else it printed.\nThose are not failed tests. Keep your own suite's processes and threads within what a command plans for.\u003C\u002Fp>\n\u003Ch3 id=\"temporary-files-and-package-stores\">Temporary files and package stores\u003C\u002Fh3>\n\u003Cp>A workspace shares its machine's disk with other runs, and the machine holds only so many files. Every command's\ntemporary files go to the run's own temporary directory: \u003Ccode>TMPDIR\u003C\u002Fcode>, \u003Ccode>TMP\u003C\u002Fcode> and \u003Ccode>TEMP\u003C\u002Fcode> point at it, so Node's\n\u003Ccode>os.tmpdir()\u003C\u002Fcode>, \u003Ccode>mktemp\u003C\u002Fcode>, Python's \u003Ccode>tempfile\u003C\u002Fcode> and most test runners' caches use it without being told. It is removed\nwith the workspace, so a test suite that leaves its temporary files behind leaves them for that run alone.\u003C\u002Fp>\n\u003Cp>pnpm's store and npm's cache are kept for your organisation and shared by its runs (\u003Ccode>npm_config_store_dir\u003C\u002Fcode> and\n\u003Ccode>npm_config_cache\u003C\u002Fcode>), so an install after the first one fetches and writes far less. They are never shared with\nanother organisation.\u003C\u002Fp>\n\u003Cp>pnpm 11 and later do not read \u003Ccode>npm_config_store_dir\u003C\u002Fcode>, so the same store is also set as \u003Ccode>pnpm_config_store_dir\u003C\u002Fcode>.\nBoth are set for you, whichever version of pnpm your repository uses: there is nothing to change.\u003C\u002Fp>\n\u003Cp>A tool that writes outside the clone and its temporary directory, to a fixed path such as \u003Ccode>\u002Ftmp\u002Fmy-cache\u003C\u002Fcode>, still\nfills the shared disk. Point it at \u003Ccode>TMPDIR\u003C\u002Fcode> instead.\u003C\u002Fp>\n\u003Ch3 id=\"commands-for-a-dependency-refresh\">Commands for a dependency refresh\u003C\u002Fh3>\n\u003Cp>Tasks that keep a repository's dependencies current call three more mise tasks. They are optional: a repository\nwithout them is not offered a refresh. Which package manager it uses, what counts as out of date and how a version\nis changed are the repository's own. The tasks name no package manager, so the same tasks work on any repository\nthat answers these commands, and a repository can cover container images or CI actions by changing its own script.\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Command\u003C\u002Fth>\n\u003Cth>Called as\u003C\u002Fth>\n\u003Cth>What it has to do\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:deps\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:deps\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Print what is out of date, in the form below. Exit 0 when the report is complete, with or without updates. Exit 2 (\u003Ccode>scan_failed\u003C\u002Fcode>) when the scan could not be completed: never print an empty report for a scan that failed.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:deps:apply\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:deps:apply [--major] &lt;name@version&gt;...\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Move each named dependency to exactly the version named, and update the lockfile. Without \u003Ccode>--major\u003C\u002Fcode>, stay on its current major. Skip one that is already there, so a line can be run again. Exit non-zero, changing nothing, when an argument matches nothing the repository declares.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:deps:verify\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:deps:verify\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Exit non-zero, naming each one, when a dependency now resolves lower than on the commit the branch started from.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>The report is markdown, because a task copies parts of it into issues as they are:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-markdown\">updates: 2 security: 1 majors: 1 held: 1\n\n## Updates to apply\n\n| Package | Current | Target | Level | Security | Note |\n| ------- | ------- | ------ | ----- | -------- | ---- |\n| `left`  | 1.0.0   | 1.0.1  | patch | high     |      |\n| `right` | 2.1.0   | 2.2.0  | minor |          |      |\n\nApply in this order, running the checks after each group:\n\n1. Security: `mise run os:deps:apply left@1.0.1`\n2. Minor: `mise run os:deps:apply right@2.2.0`\n\n## Majors\n\n### Major upgrade: framework 3 to 4\n\n- `framework` 3.2.0 to 4.1.0 (released 2026-08-25)\n- Release notes: https:\u002F\u002Fexample.com\u002Fframework\u002Freleases\n- Apply: `mise run os:deps:apply --major framework@4.1.0`\n\n## Held back\n\n- `odd` 1.0.0: the registry names no version to compare\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>What the tasks rely on:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>The first line\u003C\u002Fstrong> is the counts. \u003Ccode>updates\u003C\u002Fcode> and \u003Ccode>majors\u003C\u002Fcode> both 0 means there is nothing to do.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>\u003Ccode>## Updates to apply\u003C\u002Fcode>\u003C\u002Fstrong> holds the changes that are safe to make together: a table, then one runnable\n\u003Ccode>mise run os:deps:apply ...\u003C\u002Fcode> line for each group, numbered in the order to apply them. These become one issue,\nand one task runs the lines exactly as printed. What an argument looks like is the repository's choice.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>\u003Ccode>## Majors\u003C\u002Fcode>\u003C\u002Fstrong> has one \u003Ccode>### \u003C\u002Fcode> heading for each upgrade that may break the code, in the order to take them. The\nheading is used as the issue's title, so it must read the same on every scan for as long as that upgrade is on\noffer. A person rejects an upgrade by closing its issue as not planned, and a title that is already on an issue,\nopen or closed, is never filed again. Under the heading: what moves, where to read what changed, and its own\napply line.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>\u003Ccode>## Held back\u003C\u002Fcode>\u003C\u002Fstrong> names anything the scan could not decide, with the reason. Nothing is done with these.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The whole report\u003C\u002Fstrong> fits in 32,000 characters. When it would not, leave out the least urgent updates and say how\nmany were left out.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Deciding what is safe is the repository's job too. Things worth deciding there: how old a release must be before it\nis offered, whether a fix for a known vulnerability waits that long, and never offering a prerelease or a withdrawn\nrelease.\u003C\u002Fp>\n\u003Ch4>Runtimes are dependencies too\u003C\u002Fh4>\n\u003Cp>The runtime and the package manager (\u003Ccode>node\u003C\u002Fcode>, \u003Ccode>pnpm\u003C\u002Fcode>, \u003Ccode>npm\u003C\u002Fcode>) go through the same report, apply and verify as the\npackages, so a new Node.js release, a security release above all, reaches the repository through its dependency\nrefresh with no person having to notice it. The scanner in \u003Ccode>@juicyllama\u002Frepo\u003C\u002Fcode> 1.3.0 reads them from every tracked\nfile that declares one: a mise or \u003Ccode>.tool-versions\u003C\u002Fcode> pin, \u003Ccode>.nvmrc\u003C\u002Fcode>, \u003Ccode>engines\u003C\u002Fcode>, \u003Ccode>packageManager\u003C\u002Fcode>, \u003Ccode>FROM node:&lt;tag&gt;\u003C\u002Fcode> and\n\u003Ccode>ARG NODE_*_VERSION=\u003C\u002Fcode> in a Dockerfile, and \u003Ccode>node:&lt;tag&gt;\u003C\u002Fcode> images in compose files, workflows and Terraform.\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Every declaration's \u003Ccode>file:line\u003C\u002Fcode>\u003C\u002Fstrong> is named, in the row's note or under the major's heading, so the issue says\nexactly what moves.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>A Node security release skips the cooldown.\u003C\u002Fstrong> It is offered at once, as is a package-manager release that fixes\nan advisory on the current version.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>A floating declaration is a \u003Ccode>pin\u003C\u002Fcode> row.\u003C\u002Fstrong> \u003Ccode>node:26-alpine\u003C\u002Fcode> builds from whatever the cached tag held, so applying\nthe row writes the exact release (\u003Ccode>node:26.12.1-alpine\u003C\u002Fcode>). It counts as a security fix while its major's latest\nsecurity release is under 30 days old.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>A \u003Ccode>Security:\u003C\u002Fcode> line sits under a major\u003C\u002Fstrong> whose current line is past its end of life, or has an advisory with no\nfix left in it, and the major counts in \u003Ccode>security:\u003C\u002Fcode>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Held back\u003C\u002Fstrong>, each with its \u003Ccode>file:line\u003C\u002Fcode>: an engines floor (\u003Ccode>&gt;=22\u003C\u002Fcode>) or a range across majors, a declaration that\nnames no major (\u003Ccode>lts\u002F*\u003C\u002Fcode>), an image pinned by digest, and declarations of one runtime that disagree.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>A runtime move edits the declarations only and runs no install. A file generated from one of them is the\nrepository's own \u003Ccode>os:deps:apply\u003C\u002Fcode> task to regenerate afterwards. \u003Ccode>os:deps:verify\u003C\u002Fcode> fails on an exact runtime\ndeclaration that went down, by its file, as it does on a package.\u003C\u002Fp>\n\u003Ch3 id=\"commands-for-infrastructure\">Commands for infrastructure\u003C\u002Fh3>\n\u003Cp>A task that applies a repository's infrastructure calls two more mise tasks. They are optional: a repository\nwithout them is never asked to apply anything. What a &quot;stack&quot; is, which tool plans it and where its state lives are\nthe repository's own; the task names no tool. The credentials the commands need come from the environment the\ntask's version declares (\u003Ca href=\"#credentials-for-the-repositorys-own-commands\">Credentials for the repository's own commands\u003C\u002Fa>),\nnever from a file in the repository.\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Command\u003C\u002Fth>\n\u003Cth>Called as\u003C\u002Fth>\n\u003Cth>What it has to do\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:infra:plan\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:infra:plan &lt;stack&gt;\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Plan the stack against its real state and print what would change, under 5,000 characters, ending with \u003Ccode>plan-sha: &lt;id&gt;\u003C\u002Fcode>: a signature of the planned changes with every secret value left out. Exit 0 when nothing would change, 2 when something would, 1 on error, 9 when another run holds the state (\u003Ccode>state_locked\u003C\u002Fcode>). The summary is what the owner reads on the approval card, so name resources and the attributes that change, never a value.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>os:infra:apply\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>mise run os:infra:apply &lt;stack&gt; --plan-sha &lt;id&gt;\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Plan again and apply only while the signature still matches \u003Ccode>&lt;id&gt;\u003C\u002Fcode>; print \u003Ccode>plan_drift\u003C\u002Fcode> and exit 8 otherwise, with nothing applied, so what the owner approved is what runs. Exit 0 when applied or nothing to apply, 1 on error, 9 locked.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>The task gates the apply on exactly that command (a gate on \u003Ccode>workspace.run_command\u003C\u002Fcode> whose \u003Ccode>match\u003C\u002Fcode> names\n\u003Ccode>[&quot;mise&quot;, &quot;run&quot;, &quot;os:infra:apply&quot;]\u003C\u002Fcode>), with the plan summary as the gate summary, so the owner approves the signature\nthe apply then holds itself to. A pull request's check runs the repository's own validation of a stack with no\ncredentials and no plan; the real plan is read at the apply.\u003C\u002Fp>\n\u003Ch3 id=\"release-copy-os-release\">Release copy: \u003Ccode>os:release\u003C\u002Fcode>\u003C\u002Fh3>\n\u003Cp>Release preparation is optional, like dependency refresh. A repository without \u003Ccode>os:release\u003C\u002Fcode> opts out.\nThe repository owns its note format, audience, brand rules, version mirrors and non-shipping path list.\nThe task writes only \u003Ccode>.release\u002Fpending\u002F**\u003C\u002Fcode> on the PR branch, before merge approval. Runs still never bump\nversions, stamp dates, tag, publish or deploy.\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Task\u003C\u002Fth>\n\u003Cth>Result\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run os:release --pr N --base REF --labels JSON\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>JSON contract: notes enabled\u002Fformat\u002Fpath\u002Faudience\u002Fstyle, brand spellings, \u003Ccode>pendingPath\u003C\u002Fcode>, deterministic \u003Ccode>bump\u003C\u002Fcode>, and \u003Ccode>nonShipping\u003C\u002Fcode> for the branch diff. Exit 0 on success; unavailable context is an error, never an opt-out.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run os:release:verify --pr N --base REF --labels JSON\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Exactly one valid intent for this shipping PR, or a non-shipping diff with no intent. Exit 2 missing, 3 invalid, 4 brand lint, 5 unavailable context.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run release\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Cstrong>CI only, outside the \u003Ccode>os:*\u003C\u002Fcode> contract.\u003C\u002Fstrong> Consume all pending intents on main, bump once at the highest level, stamp each note, and print the version. No task may call it.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>\u003Ccode>N\u003C\u002Fcode> is the PR number, \u003Ccode>REF\u003C\u002Fcode> its base ref, and \u003Ccode>JSON\u003C\u002Fcode> the combined PR and linked issue label names.\nImplementations may also accept \u003Ccode>RELEASE_PR\u003C\u002Fcode>, \u003Ccode>OS_BASE_REF\u003C\u002Fcode>, and \u003Ccode>RELEASE_LABELS\u003C\u002Fcode>. \u003Ccode>major\u003C\u002Fcode> wins over\n\u003Ccode>minor\u003C\u002Fcode>; absent either label means \u003Ccode>patch\u003C\u002Fcode>. The model never chooses the bump.\u003C\u002Fp>\n\u003Cp>A shipping PR writes \u003Ccode>.release\u002Fpending\u002Fpr-N.md\u003C\u002Fcode>:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-markdown\">---\nbump: patch\ntitle: Wallet sheets survive focus changes\ndescription: Buyers can return to checkout without restarting payment.\ntags: [Fix]\n---\n\nThe customer-facing change and any integration action, in this repository's public voice.\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>There is no version or date. When notes are disabled, metadata remains and the body is empty.\nThe filename is stable across retries. Existing pending files inherited from main are left alone.\nA branch changing only declared non-shipping paths needs no intent. Every other merge releases.\u003C\u002Fp>\n\u003Cp>The note-writing task reads the PR, linked issue and diff, checks the contract, writes the intent and\nverifies before committing. The merge approval includes the bump and title so the approver sees the copy.\nA PR check enforces the same rules for manually written PRs. Enable that check only after the writer is\navailable. A changed PR or linked issue label needs fresh verification before approval.\u003C\u002Fp>\n\u003Cp>The repository's release workflow releases only a commit that passed its checks, serializes releases, pushes the\nversioned main commit before its tag, and then publishes or deploys that commit. Several merged intents\ncan share one release version; each retains its own note. No pending intents means no release.\nThe public \u003Ccode>@juicyllama\u002Frepo\u003C\u002Fcode> package (1.3.0) provides the Node implementation and reusable GitHub workflows, as\nwell as the dependency scanner and the test-report wrapper above. Run them from the one installed version\n(\u003Ccode>node node_modules\u002F@juicyllama\u002Frepo\u002F...\u003C\u002Fcode>) rather than fetching a second copy, so a bump of the package moves them\nall. Other stacks may implement the same contract. Historical release gaps are not backfilled.\u003C\u002Fp>\n\u003Ch3 id=\"public-changelog-os-changelog\">Public changelog: \u003Ccode>os:changelog\u003C\u002Fcode>\u003C\u002Fh3>\n\u003Cp>A public changelog is optional too, and separate from release copy: a repository can implement either without the\nother. An enterprise that publishes stories about its finished work names the repository that hosts them once, in\nthe \u003Cstrong>Public changelog repository\u003C\u002Fstrong> setting on its Enterprise page. Shared tasks never name it, so updating one never\nchanges it. Every run carries it in its context as \u003Ccode>enterprise.changelogRepo\u003C\u002Fcode>, \u003Ccode>null\u003C\u002Fcode> when none is named. It is\nnamed as \u003Ccode>owner\u002Fname\u003C\u002Fcode> or a clone address without credentials: access to a private repository comes from the code\nhost access the enterprise has bound, never from the address.\u003C\u002Fp>\n\u003Cp>That repository says where a story goes, what it must look like and how it is checked. Its paths, story format,\naudience, voice, brand spellings and image rules are its own; the tasks that write a story name none of them.\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Task\u003C\u002Fth>\n\u003Cth>Result\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run os:changelog --epic REF --title TEXT\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Exit 0 and print the JSON contract below. Any other exit means unavailable context: an error, never an opt-out.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run os:changelog:verify --epic REF --story PATH\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Check one story against the repository's own rules before its pull request is opened, with the exits below.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run os:changelog:image --epic REF --story PATH\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Optional. Make the hero image for a story already written at \u003Ccode>PATH\u003C\u002Fcode>, with the exits below.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>mise run os:changelog:release-notes --epic REF --merged OWNER\u002FNAME#N ...\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Optional. Give back the public release-note pages for the changes a story was written from, with the exits below.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>\u003Ccode>REF\u003C\u002Fcode> is the epic the story is about, as \u003Ccode>owner\u002Frepo#number\u003C\u002Fcode>, and \u003Ccode>TEXT\u003C\u002Fcode> is that epic's title.\n\u003Ccode>PATH\u003C\u002Fcode> is the story file, relative to the repository root.\nEach \u003Ccode>OWNER\u002FNAME#N\u003C\u002Fcode> is one change the story was written from: a merged pull request, in the repository it merged in.\nThe same inputs on the same commit always give the same answer.\nSo the story's path stays the same across retries, and a retried run finds the story it already wrote.\u003C\u002Fp>\n\u003Cp>\u003Ccode>os:changelog\u003C\u002Fcode> prints one JSON object:\u003C\u002Fp>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Field\u003C\u002Fth>\n\u003Cth>What it holds\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>enabled\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>false\u003C\u002Fcode> when this repository hosts no public changelog. The other fields may then be left out.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>path\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Where this epic's story goes, relative to the repository root. Decided only by \u003Ccode>--epic\u003C\u002Fcode>, \u003Ccode>--title\u003C\u002Fcode> and the tree.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>format\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The story's type, the front-matter fields it needs and what each holds, and the rules for its body.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>audience\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Who reads the story.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>style\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The voice, tense and length to write in.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>brand\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Each correct spelling, with the wrong variants a story may not use.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>neverName\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>What a story must never name, on top of the rule every story keeps: no task names, issue numbers, members or unreleased work.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>links\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Optional. The public URL prefixes a story may link to. When present, a story that links outside them fails the check with exit 4. When absent, links are left to the writing rules.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>learnMore\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Optional. The public pages a story's &quot;Learn more&quot; section may list, each with a \u003Ccode>url\u003C\u002Fcode>, a \u003Ccode>title\u003C\u002Fcode> and an \u003Ccode>about\u003C\u002Fcode> that says what the page covers.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>images\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Optional. The rules for images: where they go (\u003Ccode>path\u003C\u002Fcode>), which formats and sizes, and the alt text each needs. \u003Ccode>hero\u003C\u002Fcode>, when present, is the hero image's \u003Ccode>width\u003C\u002Fcode>, \u003Ccode>height\u003C\u002Fcode> and \u003Ccode>formats\u003C\u002Fcode>; a hero is a \u003Ccode>png\u003C\u002Fcode>, \u003Ccode>webp\u003C\u002Fcode>, \u003Ccode>jpg\u003C\u002Fcode> or \u003Ccode>jpeg\u003C\u002Fcode> file, the only image files a story pull request may carry.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cpre>\u003Ccode class=\"language-json\">{\n    &quot;enabled&quot;: true,\n    &quot;path&quot;: &quot;changelog\u002Fexport-reports.md&quot;,\n    &quot;format&quot;: {\n        &quot;type&quot;: &quot;markdown&quot;,\n        &quot;frontMatter&quot;: {\n            &quot;title&quot;: &quot;The story's headline&quot;,\n            &quot;date&quot;: &quot;Publication date, YYYY-MM-DD&quot;,\n            &quot;tags&quot;: &quot;One or more of Feature, Fix&quot;,\n            &quot;image&quot;: &quot;The hero image's path, from os:changelog:image&quot;,\n            &quot;imageAlt&quot;: &quot;The hero image's alt text&quot;\n        },\n        &quot;body&quot;: &quot;Two to four short paragraphs, then an optional Learn more section. No other headings.&quot;\n    },\n    &quot;audience&quot;: &quot;People who use the product, not the team that builds it&quot;,\n    &quot;style&quot;: &quot;Plain, present tense, second person&quot;,\n    &quot;brand&quot;: { &quot;GitHub&quot;: [&quot;Github&quot;, &quot;github&quot;] },\n    &quot;neverName&quot;: [&quot;Customer names&quot;, &quot;Internal service names&quot;],\n    &quot;links&quot;: [&quot;https:\u002F\u002Fexample.com\u002Fdocs\u002F&quot;],\n    &quot;learnMore&quot;: [\n        {\n            &quot;url&quot;: &quot;https:\u002F\u002Fexample.com\u002Fdocs\u002Freports\u002Fexport\u002F&quot;,\n            &quot;title&quot;: &quot;Export a report&quot;,\n            &quot;about&quot;: &quot;Exporting reports as CSV or PDF&quot;\n        },\n        {\n            &quot;url&quot;: &quot;https:\u002F\u002Fexample.com\u002Fdocs\u002Fbilling\u002F&quot;,\n            &quot;title&quot;: &quot;Billing&quot;,\n            &quot;about&quot;: &quot;Plans, invoices and payment methods&quot;\n        }\n    ],\n    &quot;images&quot;: {\n        &quot;path&quot;: &quot;changelog\u002Fimages\u002F&quot;,\n        &quot;formats&quot;: [&quot;png&quot;, &quot;webp&quot;],\n        &quot;maxWidth&quot;: 1600,\n        &quot;alt&quot;: &quot;required&quot;,\n        &quot;hero&quot;: { &quot;width&quot;: 1600, &quot;height&quot;: 900, &quot;formats&quot;: [&quot;png&quot;] }\n    }\n}\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>\u003Cstrong>Learn more.\u003C\u002Fstrong> Every \u003Ccode>url\u003C\u002Fcode> in \u003Ccode>learnMore\u003C\u002Fcode> is a public page, and when \u003Ccode>links\u003C\u002Fcode> is present it sits inside one of the \u003Ccode>links\u003C\u002Fcode> prefixes.\nA story's &quot;Learn more&quot; section, at its end, lists only \u003Ccode>learnMore\u003C\u002Fcode> entries whose \u003Ccode>about\u003C\u002Fcode> relates to the feature the story is about, each linked with its \u003Ccode>title\u003C\u002Fcode>.\nThe section is left out when none of them relate, or when \u003Ccode>learnMore\u003C\u002Fcode> is absent.\nA prefix in \u003Ccode>links\u003C\u002Fcode> allows a link; an entry in \u003Ccode>learnMore\u003C\u002Fcode> names one page and says what it covers, so a story never invents a page to point to.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>Hero image.\u003C\u002Fstrong> \u003Ccode>os:changelog:image\u003C\u002Fcode> is optional: it makes the story's hero image, in the repository's own brand, once the story is written at \u003Ccode>PATH\u003C\u002Fcode>.\nIt writes the image under \u003Ccode>images.path\u003C\u002Fcode>, at the \u003Ccode>width\u003C\u002Fcode> and \u003Ccode>height\u003C\u002Fcode> and in one of the \u003Ccode>formats\u003C\u002Fcode> that \u003Ccode>images.hero\u003C\u002Fcode> declares, on an opaque background, so it reads on light and dark pages and in link previews.\nIt prints one JSON object, \u003Ccode>{ &quot;path&quot;, &quot;alt&quot;, &quot;text&quot; }\u003C\u002Fcode>: \u003Ccode>path\u003C\u002Fcode> is the image file, relative to the repository root, \u003Ccode>alt\u003C\u002Fcode> is its alt text, and \u003Ccode>text\u003C\u002Fcode> is every word drawn on the image, taken only from the story's own title and summary.\nSo the image names nothing the story does not, and cannot contradict it.\nIt makes the same image for the same story on the same commit.\nThe story carries \u003Ccode>path\u003C\u002Fcode> and \u003Ccode>alt\u003C\u002Fcode> in the front-matter fields \u003Ccode>format\u003C\u002Fcode> names (\u003Ccode>image\u003C\u002Fcode> and \u003Ccode>imageAlt\u003C\u002Fcode> in the example above), so the changelog shows the image and uses it when the story is shared.\nA hero is a \u003Ccode>png\u003C\u002Fcode>, \u003Ccode>webp\u003C\u002Fcode>, \u003Ccode>jpg\u003C\u002Fcode> or \u003Ccode>jpeg\u003C\u002Fcode> file: a hero declared in any other format is reported as an image error, and the story goes out text-only.\u003C\u002Fp>\n\u003Cp>\u003Ccode>os:changelog:image\u003C\u002Fcode> exits:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>0\u003C\u002Fstrong> made the image: the story uses it as its hero.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>2\u003C\u002Fstrong> no image for this story: the story goes out text-only.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Any other exit is an error: the story still goes out text-only, and the error is reported with it, so a broken command stays visible instead of reading as no image.\nA repository that does not define \u003Ccode>os:changelog:image\u003C\u002Fcode> has no generated hero images, and its stories go out text-only.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>Release notes.\u003C\u002Fstrong> \u003Ccode>os:changelog:release-notes\u003C\u002Fcode> is optional: it gives back the public pages of the release notes the repository publishes for the changes a story was written from, so a reader who wants the detail behind the story can find it.\nThe OS passes the epic as \u003Ccode>--epic\u003C\u002Fcode> and one \u003Ccode>--merged\u003C\u002Fcode> for each merged pull request the story was written from, and nothing for work still open, closed as not planned or closed unmerged.\nPublic release notes carry no internal references, so the match between a note and its change happens inside the repository, which knows how it releases them.\nIt prints one JSON list, \u003Ccode>[{ &quot;url&quot;, &quot;title&quot; }]\u003C\u002Fcode>: each \u003Ccode>url\u003C\u002Fcode> is a public page, \u003Ccode>http\u003C\u002Fcode> or \u003Ccode>https\u003C\u002Fcode> with no sign-in details and, when \u003Ccode>links\u003C\u002Fcode> is present, inside one of its prefixes, and each \u003Ccode>title\u003C\u002Fcode> is the note's own title.\nNothing internal comes back: no pull request numbers, no \u003Ccode>owner\u002Fname#N\u003C\u002Fcode> references and no file paths.\nIt gives the same answer for the same inputs on the same commit.\nThe story lists the pages in a &quot;Release notes&quot; section, before &quot;Learn more&quot; when it has both, each linked under its own title, once, in the order the changes merged.\nAn entry that is not a public page, or whose title or \u003Ccode>url\u003C\u002Fcode> names a pull request or issue number, an \u003Ccode>owner\u002Fname#N\u003C\u002Fcode> reference or a file path, is left out, never edited to pass, and the &quot;Release notes&quot; section is left out when no page is left: a story never has an empty heading.\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-json\">[\n    { &quot;url&quot;: &quot;https:\u002F\u002Fexample.com\u002Fdocs\u002Freleases\u002Fexport-reports-as-pdf\u002F&quot;, &quot;title&quot;: &quot;Export reports as PDF&quot; },\n    { &quot;url&quot;: &quot;https:\u002F\u002Fexample.com\u002Fdocs\u002Freleases\u002Fschedule-a-report\u002F&quot;, &quot;title&quot;: &quot;Schedule a report&quot; }\n]\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>\u003Ccode>os:changelog:release-notes\u003C\u002Fcode> exits:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>0\u003C\u002Fstrong> the pages: the story links them in its &quot;Release notes&quot; section.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>2\u003C\u002Fstrong> no published notes for these changes: the story has no &quot;Release notes&quot; section.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Any other exit is an error, and so is output that is not the JSON list: the story still goes out, with no &quot;Release notes&quot; section, and the problem is named on a &quot;Release notes:&quot; line of its pull request description, so a broken command stays visible instead of reading as no notes.\nA failing release-notes command never stops a story.\nA repository that does not define \u003Ccode>os:changelog:release-notes\u003C\u002Fcode> publishes no release notes to link, so its stories have no &quot;Release notes&quot; section and are otherwise unchanged.\u003C\u002Fp>\n\u003Cp>\u003Ccode>os:changelog:verify\u003C\u002Fcode> exits:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>0\u003C\u002Fstrong> valid: the story can be committed and its pull request opened.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>2\u003C\u002Fstrong> story missing: there is no file at \u003Ccode>PATH\u003C\u002Fcode>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>3\u003C\u002Fstrong> invalid format: the story is not at \u003Ccode>path\u003C\u002Fcode>, its front matter or body breaks \u003Ccode>format\u003C\u002Fcode>, or it has a &quot;Learn more&quot; or &quot;Release notes&quot; heading with no links under it.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>4\u003C\u002Fstrong> public-writing lint: a wrong brand spelling, something \u003Ccode>neverName\u003C\u002Fcode> or the rule every story keeps forbids, a link outside \u003Ccode>links\u003C\u002Fcode>, a &quot;Learn more&quot; link that is not in \u003Ccode>learnMore\u003C\u002Fcode>, a &quot;Release notes&quot; link that \u003Ccode>os:changelog:release-notes\u003C\u002Fcode> does not return for the epic's merged pull requests, an image that breaks \u003Ccode>images\u003C\u002Fcode>, or an image whose drawn text does not come from the story's title or summary. Verify gets no \u003Ccode>--merged\u003C\u002Fcode> list, so for a &quot;Release notes&quot; link it reads every merged pull request of the epic \u003Ccode>--epic\u003C\u002Fcode> names itself, and checks the link against the pages \u003Ccode>os:changelog:release-notes\u003C\u002Fcode> returns for all of them: a page the command returned for the pull requests a story was written from, which are some of that set, must pass.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>5\u003C\u002Fstrong> unavailable context: the epic, the tree or the rules could not be read. This is never a valid story.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>A story with no image is valid (exit 0), whether or not the repository defines \u003Ccode>os:changelog:image\u003C\u002Fcode>.\u003C\u002Fp>\n\u003Cp>A failed check prints what is wrong, one line per problem, as \u003Ccode>&lt;rule&gt;: &lt;file&gt;:&lt;line&gt;: &lt;what&gt;\u003C\u002Fcode>, and exits with the\ncode above. When a story breaks both its format and the public-writing rules, every problem is printed and it exits 3.\nA &quot;Learn more&quot; link that is not in \u003Ccode>learnMore\u003C\u002Fcode> prints \u003Ccode>learnMore: &lt;file&gt;:&lt;line&gt;: &lt;url&gt; is not in learnMore\u003C\u002Fcode>.\nDrawn image text that does not come from the story's title or summary is reported under \u003Ccode>images\u003C\u002Fcode>.\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-text\">brand: changelog\u002Fexport-reports.md:7: &quot;Github&quot; is spelled &quot;GitHub&quot;\nlinks: changelog\u002Fexport-reports.md:12: https:\u002F\u002Finternal.example.net\u002Freport is outside the declared links\nlearnMore: changelog\u002Fexport-reports.md:18: https:\u002F\u002Fexample.com\u002Fdocs\u002Fapi\u002F is not in learnMore\nimages: changelog\u002Fexport-reports.md:4: the hero's drawn text &quot;Beta for Acme&quot; is not in the story's title or summary\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>An enterprise has no public changelog when any of these holds:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>it names no repository in \u003Cstrong>Public changelog repository\u003C\u002Fstrong>;\u003C\u002Fli>\n\u003Cli>the named repository does not define an \u003Ccode>os:changelog\u003C\u002Fcode> task;\u003C\u002Fli>\n\u003Cli>\u003Ccode>os:changelog\u003C\u002Fcode> prints \u003Ccode>&quot;enabled&quot;: false\u003C\u002Fcode>;\u003C\u002Fli>\n\u003Cli>the repository cannot be read with the code host access the enterprise has bound.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Then nothing fails and nothing is retried: a task that would write a story finishes without one.\nA defined \u003Ccode>os:changelog\u003C\u002Fcode> that exits non-zero, or prints anything but this contract, is an error, never an opt-out.\nIt is reported as an error, so a broken implementation stays visible instead of reading as no changelog.\u003C\u002Fp>\n\u003Ch2 id=\"llm-instruction-for-migrating-an-existing-workspace\">LLM Instruction for migrating an existing workspace\u003C\u002Fh2>\n\u003Cp>Open the existing repository in a coding assistant that can read and edit files and run commands, then paste\nthe instruction below. It asks the assistant to implement and validate the integration using this page's\ncontracts. You keep the repository, its stack and its CI; the migration adds the commands the engineering team\nneeds. If the assistant cannot open this page, give it the page's contents before it starts.\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-text\">Onboard this existing development repository as a Zero Human workspace so a cloned\nengineering team can work on it. Implement the integration in this repository.\n\nRead https:\u002F\u002Fzerohuman.com\u002Fdocs\u002Ftools\u002Fworkspaces in full, including the setup,\nverification, dependency and release contracts. Use that page as the contract.\nIf you cannot access it, ask me for its contents instead of guessing.\nRead this repository's AGENTS.md and contributor\u002Fdevelopment instructions first.\n\n1. Inspect before editing.\n   Identify the language\u002Fruntime versions, package manager and lockfiles, monorepo\n   layout, test frameworks and locations, lint\u002Ftypecheck\u002Fbuild commands, git hooks,\n   CI requirements, coverage thresholds, release process and existing mise config.\n   Detect the remote's default branch. Identify required generated files, test\n   services and private dependencies. Preserve existing work and use an isolated\n   branch or worktree. Summarize the mapping from existing commands to the hooks\n   below, then implement it. Ask only for decisions you cannot infer safely.\n\n2. Make setup work from a clean, non-interactive clone.\n   Add to mise.toml rather than replacing existing configuration. Keep the current\n   stack and package manager; pin compatible runtimes where needed. Use the existing\n   lockfile and reproducible installs. Add .os\u002Fworkspace.yml only when auto-detection\n   is insufficient, or update the existing manifest without creating a competing one.\n   Explicit setup replaces ALL auto-detection: include tool installation, dependency\n   installation, code generation and hook installation as needed.\n   Each setup\u002Fteardown step is an argv array starting with git, node, npm, pnpm or\n   mise; there are at most eight steps in each list and 16 argv entries per step.\n   Put shell logic in repo scripts or mise tasks. Do not put shell strings, pipes,\n   redirects or cd in manifest argv. Validate mise trust and runtime activation in\n   a fresh home directory; do not depend on my shell profile or global packages.\n   Setup must work with CI=true and NODE_ENV=test. Do not copy secrets into the repo.\n   Document missing credentials or services instead of assuming they are available.\n\n3. Implement the three core mise hooks with real checks.\n   - mise run os:red &lt;issue&gt;: run the test files this branch adds or changes relative\n     to the default branch, including uncommitted tests, and print every suite path.\n     Use normal test filenames, not issue-number filenames. Its description must\n     tell an agent where tests belong and which framework and conventions to use.\n     Exit 0 for green, 1 for assertion failures, 3 for no suite, 4 for typecheck\n     failures, 5 for skipped\u002Ffocused\u002Ftodo tests. Setup\u002Fimport errors are not RED proof.\n   - mise run os:check [issue]: run lint, typecheck, tests and prerequisite builds.\n     Without an issue, check the whole repository. Any scoped run must cover affected\n     code and dependencies. Add --parts and --part &lt;name&gt; if the full check exceeds\n     ten minutes; every part must fit and a part checks the repo without an issue.\n     --parts accepts an optional issue and may then print only the parts it needs.\n   - mise run os:coverage [issue]: enforce the project's thresholds on changed source\n     lines, print uncovered changed lines per file, and fail below the thresholds.\n     Report explicitly when there is no measurable source change. If no coverage\n     policy exists, propose one and ask me to choose it; do not silently pick zero.\n     Add --parts and --part &lt;name&gt; if measuring exceeds ten minutes: every part must\n     fit, the last part prints the verdict, and it must refuse stale or missing parts.\n   Forward arguments and exit codes through every wrapper. Resolve the real default\n   branch and merge base, fetching more history when the 50-commit clone is too shallow.\n   Missing history must fail clearly. Keep each invocation within ten minutes and\n   32,000 output characters: print a verdict and failure details, save full logs to\n   ignored files. Bound concurrency for CI resources. Only implement diagnostic\n   exits 6 and 7 if you can prove the conditions in the workspace guide.\n\n4. Preserve the RED-to-GREEN workflow and merge checks.\n   Git hooks must install and run normally. Deliberate assertion failures in the\n   RED phase must be committable and pushable to a working branch, with static\n   checks still enforced. If hooks or branch rules currently prevent that, identify\n   the conflict and make the smallest explicit adjustment permitted by repo policy.\n   Keep full tests, coverage and required CI checks before merge. Never use\n   --no-verify, HUSKY=0, skipped tests, reduced thresholds or success-only stubs.\n\n5. Assess optional hooks; implement only what the project will use.\n   For dependency maintenance, implement os:deps, os:deps:apply and os:deps:verify\n   together, matching the report, exact-version, --major and downgrade contracts.\n   For release-note preparation, assess os:release and os:release:verify against the\n   existing release process. If adopted, implement the JSON contract and pending\n   intents from the guide and keep versioning\u002Ftagging\u002Fpublishing\u002Fdeployment in CI.\n   Do not replace the project's release process without agreement. Otherwise list\n   these optional capabilities as not configured; do not install empty placeholders.\n   Do not invent other mandatory os:* hooks. A helper such as os:setup only runs\n   when the manifest or another command explicitly invokes it.\n\n6. Validate the integration, not just the file syntax.\n   In a disposable fresh clone with a clean home and CI-like environment, run the\n   actual setup, discover the mise tasks and run every implemented hook. Exercise\n   argument forwarding, os:red's no-suite case, a temporary assertion failure and\n   its passing counterpart, typecheck failure where applicable, and skip\u002Ffocus\u002Ftodo\n   rejection. Confirm coverage fails below its threshold and handles no changed\n   source. Check every os:check part if implemented. Exercise local git hooks on a\n   disposable branch without pushing. Remove temporary probes afterwards.\n   Report commands, exit codes, timings and any checks blocked by this environment.\n   Do not claim a hosted Zero Human run has passed from local checks alone.\n\n7. Leave a reviewable change and an onboarding handoff.\n   Document the files added\u002Fchanged, hook commands and descriptions, prerequisites,\n   coverage policy, optional capabilities and remaining blockers in the repo's\n   developer docs. Explain the remaining Zero Human setup: clone the engineering\n   team, connect a code-host MCP tool with repository access, and configure git.host\n   and git.username. Set git.repoTemplate for short owner\u002Fname inputs or use a full\n   HTTPS repo input; optionally set git.repo. Check task write globs\u002Frequired files\n   against this repo's paths.\n   Cloned tasks may require their source owner's changes or an independently owned\n   copy; repo config does not override task permissions. Keep credentials and team\n   configuration out of committed files. Leave merge\u002Fdeploy and the first hosted\n   smoke run for me to approve. Finish with what is ready and what remains unverified.\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Ch3 id=\"connect-the-team-after-the-repository-change\">Connect the team after the repository change\u003C\u002Fh3>\n\u003Cp>Merge the reviewed onboarding files before starting new work, so fresh branches inherit the setup. Cloning a\nteam does not configure repository credentials or make its task write rules fit every project. Connect the\n\u003Ca href=\"#the-credential-comes-from-the-code-host-tool-you-bind\">code host binding\u003C\u002Fa>, check the team's task paths and\nrequired files, and configure its agents before the first run. Workspace support for a git host does not\nautomatically translate a cloned task's GitHub issue or pull-request tools to another host's tools.\u003C\u002Fp>\n\u003Cp>Start with a small issue on a disposable working branch. Confirm setup succeeds, the expected tests run, the\nmember commits and pushes, and the pull request reaches the configured merge approval with the expected CI\nchecks. Local validation proves the repository integration; this first run proves the team's bindings and\nworkflow work with it.\u003C\u002Fp>\n\u003Ch2 id=\"what-a-task-decides\">What a task decides\u003C\u002Fh2>\n\u003Cp>A task version's \u003Cstrong>Workspace\u003C\u002Fstrong> decides which branch its runs work on and which files they may change. On a task\nin the portal it is \u003Cstrong>Workspace JSON\u003C\u002Fstrong>:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-json\">{\n    &quot;branch&quot;: &quot;fix\u002F{issue}&quot;,\n    &quot;writeAllow&quot;: [&quot;src\u002F**&quot;, &quot;tests\u002F**&quot;],\n    &quot;writeDeny&quot;: [&quot;src\u002Fgenerated\u002F**&quot;]\n}\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Field\u003C\u002Fth>\n\u003Cth>What it does\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>branch\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The branch template (see \u003Ca href=\"#the-branch\">The branch\u003C\u002Fa>).\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>onDefaultBranch\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>true\u003C\u002Fcode> keeps the run on the default branch's tip with no branch of its own and nothing to push (see \u003Ca href=\"#the-branch\">The branch\u003C\u002Fa>).\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>pushAllow\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The full ref globs (\u003Ccode>refs\u002Fheads\u002Fmain\u003C\u002Fcode>, \u003Ccode>refs\u002Ftags\u002Fv*\u003C\u002Fcode>, \u003Ccode>refs\u002Fheads\u002Fproduction\u003C\u002Fcode>) a run may push with \u003Ccode>workspace.git_push\u003C\u002Fcode>'s \u003Ccode>ref\u003C\u002Fcode> besides its own branch. It is the one way a run pushes the default branch, a tag, or a branch a host builds from, never by force, and it needs a \u003Ccode>writeAllow\u003C\u002Fcode>. For a task that cuts a release.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>writeAllow\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Paths the run may change freely. A non-empty list refuses other paths unless \u003Ccode>rewriteAllow\u003C\u002Fcode> or \u003Ccode>appendOnly\u003C\u002Fcode> matches them. Matching paths may be rewritten even if \u003Ccode>appendOnly\u003C\u002Fcode> also matches. A glob may name a run input as \u003Ccode>branch\u003C\u002Fcode> does: \u003Ccode>{contentDir}\u002F**\u003C\u002Fcode> is the folder the run's \u003Ccode>contentDir\u003C\u002Fcode> input names, so a task cloned from the catalog holds each company's runs to the folders that company set; a glob whose input is empty on a run is left out of that run. The same holds for \u003Ccode>rewriteAllow\u003C\u002Fcode>, \u003Ccode>writeDeny\u003C\u002Fcode> and \u003Ccode>appendOnly\u003C\u002Fcode>.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>rewriteAllow\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Paths the run may change freely, even where \u003Ccode>appendOnly\u003C\u002Fcode> matches, without refusing everything else.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>writeDeny\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Paths the run may never change. It wins over every other field.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>appendOnly\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Paths the run may add files and lines to, but where no line that was there when the run started may change or go.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>requireFiles\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Paths that must exist in the clone before the run starts work; each \u003Ccode>{key}\u003C\u002Fcode> is filled from the run's input. If one is missing, the run fails \u003Ccode>workspace_file_missing\u003C\u002Fcode>.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>credentials\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The values the run's repository commands need, each \u003Ccode>{ tool, as, prop? }\u003C\u002Fcode>: a bound tool's secret, or its custom prop, as the environment variable \u003Ccode>as\u003C\u002Fcode> of \u003Ccode>workspace.run_command\u003C\u002Fcode> alone (see \u003Ca href=\"#credentials-for-the-repositorys-own-commands\">Credentials for the repository's own commands\u003C\u002Fa>).\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>Paths are globs: \u003Ccode>*\u003C\u002Fcode> stays inside one folder, \u003Ccode>**\u003C\u002Fcode> crosses folders. A task with none of the path fields set may\nchange anything in the clone.\u003C\u002Fp>\n\u003Cp>The rules are checked when a file is written, edited or deleted, again on what a commit would include (a command\ncan change files too), and, for \u003Ccode>appendOnly\u003C\u002Fcode>, once more on the run's own commits when it pushes. A refused write\nchanges nothing; a refused commit commits nothing. During a merge, a file the merge brings in may be written and\ncommitted as the merge has it, whatever the rules say, so a run can finish a merge; a line of the run's own in\nsuch a file is still refused.\u003C\u002Fp>\n\u003Ch2 id=\"commits-and-pushes\">Commits and pushes\u003C\u002Fh2>\n\u003Cp>Commits are authored and committed as \u003Cstrong>the member the run belongs to\u003C\u002Fstrong>: its name and the email address on its\nmember page. A run with no member commits as the enterprise itself, under your enterprise's name, from an \u003Ccode>os@\u003C\u002Fcode>\naddress on your enterprise's Zero Human mail domain.\u003C\u002Fp>\n\u003Cp>Code hosts link a commit to an account by its email address. For commits to show under the member's own account\non your host, add the member's address to that account and verify it (for GitHub, see\n\u003Ca href=\"\u002Fdocs\u002Ftools\u002Fgithub\">GitHub\u003C\u002Fa>). The push itself is made by whichever account the credential belongs to.\u003C\u002Fp>\n\u003Cp>\u003Ccode>workspace.git_commit\u003C\u002Fcode> stages what the run changed and commits it; a new file that only a command created may need\na \u003Ccode>git add\u003C\u002Fcode> first. \u003Ccode>workspace.git_push\u003C\u002Fcode> pushes the branch to the repository's remote under the same name, as an\nordinary push: it never overwrites work someone else pushed.\u003C\u002Fp>\n\u003Cp>If a run of a task that lists \u003Ccode>workspace.git_push\u003C\u002Fcode> finishes without a successful push while its workspace holds\nchanges or commits the remote does not have, it fails: \u003Ccode>git_push_missing\u003C\u002Fcode>, or the reason the commit or push was\nrefused. Work is not lost behind a success. If a run runs out of steps, or its model call fails, the OS commits and\npushes what it had before the run fails, where the task's rules allow, so a retry picks up from there. A task that\ngates its own push gets no such save (see \u003Ca href=\"#what-waits-for-you\">What waits for you\u003C\u002Fa>).\u003C\u002Fp>\n\u003Ch2 id=\"what-is-never-allowed\">What is never allowed\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Cstrong>Pushing the default branch.\u003C\u002Fstrong> \u003Ccode>workspace.git_push\u003C\u002Fcode> refuses the repository's default branch, and any branch\nnamed \u003Ccode>main\u003C\u002Fcode> or \u003Ccode>master\u003C\u002Fcode> (\u003Ccode>default_branch_refused\u003C\u002Fcode>). A member has no other way to push: \u003Ccode>workspace.run_command\u003C\u002Fcode>\nrefuses a command with \u003Ccode>git push\u003C\u002Fcode> anywhere in it, including through \u003Ccode>mise\u003C\u002Fcode>, \u003Ccode>pnpm\u003C\u002Fcode> or \u003Ccode>node\u003C\u002Fcode>\n(\u003Ccode>git_push_via_run_command\u003C\u002Fcode>).\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Skipping the repository's hooks.\u003C\u002Fstrong> A git command with \u003Ccode>--no-verify\u003C\u002Fcode>, \u003Ccode>git commit -n\u003C\u002Fcode>, or a \u003Ccode>core.hooksPath\u003C\u002Fcode>\noverride is refused (\u003Ccode>git_hooks_skip_refused\u003C\u002Fcode>), and the file tools write nothing under \u003Ccode>.git\u002Fhooks\u003C\u002Fcode>.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>A shell.\u003C\u002Fstrong> \u003Ccode>workspace.run_command\u003C\u002Fcode> runs one binary with its arguments: no pipes, no redirects, no \u003Ccode>cd\u003C\u002Fcode>, no\n\u003Ccode>&amp;&amp;\u003C\u002Fcode>. The binary is one of \u003Ccode>git\u003C\u002Fcode>, \u003Ccode>node\u003C\u002Fcode>, \u003Ccode>npm\u003C\u002Fcode>, \u003Ccode>pnpm\u003C\u002Fcode> and \u003Ccode>mise\u003C\u002Fcode>, named, not given as a path\n(\u003Ccode>bin_not_allowlisted\u003C\u002Fcode>).\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Files outside the clone.\u003C\u002Fstrong> Every path the file tools read or write is inside the clone. A path that resolves\noutside it is refused.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2 id=\"what-waits-for-you\">What waits for you\u003C\u002Fh2>\n\u003Cp>A run's workspace waits for you only where its task says so. Without a gate, every write, commit and push goes ahead\nas the task's rules allow, and only ever to a branch that is not the default one. A gate the task declares on a\n\u003Ccode>workspace.*\u003C\u002Fcode> or \u003Ccode>memory.*\u003C\u002Fcode> tool waits like any other gated call: the OS stops the call, and runs exactly what you\napproved. The gate can name one tool, all the workspace or all the memory tools (a row naming just \u003Ccode>workspace\u003C\u002Fcode> or\n\u003Ccode>memory\u003C\u002Fcode>), or only some calls of a tool (\u003Ccode>match\u003C\u002Fcode>). A \u003Ccode>memory.*\u003C\u002Fcode> call is checked against the memory grants first, so\none they refuse fails without asking you. See \u003Ca href=\"\u002Fdocs\u002Ftools\u002Fbuilt-in#gates-and-grants\">Gates and grants\u003C\u002Fa> in the\nbuilt-in tools reference.\u003C\u002Fp>\n\u003Cp>A task that gates \u003Ccode>workspace.git_push\u003C\u002Fcode>, or all the workspace tools, gets no OS checkpoint. A checkpoint is the OS\nsaving a run's work for it: it commits the work and pushes it to the run's branch, so nothing is lost if the worker\nrunning it goes. The OS makes one when the run waits at a gate, when a release hands the run to another worker, and\nwhen the run stops because its model gave out or ran out of steps. While the push is gated, a checkpoint would send\nthe remote commits you have not approved, so the OS saves nothing on the task's behalf: the work stays only on the\nworker that holds it, and the run's own push waits for you.\u003C\u002Fp>\n\u003Cp>Gating the push has a cost. Work that is not pushed is lost if its worker goes before the run pushes it: a release,\na crash, or a run that waits at a gate for longer than the worker lives. The run then carries on on another worker,\nfrom what was last pushed. When a release handed it on, or a gate it waited at is decided there, it is told which\nfiles were lost and has to make those changes again, and it cannot finish as a success until it has pushed again,\nwhich waits for you like any other push. After a crash it is told only that unpushed work is gone. So gate the push\nonly when a push itself needs your approval; to approve what reaches the default branch, gate the merge instead\n(below).\u003C\u002Fp>\n\u003Cp>A gate limited by \u003Ccode>match\u003C\u002Fcode> to calls the OS's own push never makes does not stop the checkpoint. For example, a gate on\n\u003Ccode>workspace.git_push\u003C\u002Fcode> with \u003Ccode>&quot;match&quot;: { &quot;ref&quot;: &quot;refs\u002Fheads\u002Fproduction&quot; }\u003C\u002Fcode> holds a run's pushes to \u003Ccode>production\u003C\u002Fcode>, but the\ncheckpoint pushes the run's own branch with no \u003Ccode>ref\u003C\u002Fcode>, so it still goes ahead.\u003C\u002Fp>\n\u003Cp>The decision that waits is the merge. Getting a branch into the default branch goes through your code host: a pull\nrequest, or merge request, and the host's merge tool, which a task that merges should gate. On GitHub it has to: a\ntask that names \u003Ccode>github.merge_pull_request\u003C\u002Fcode> cannot be saved without a gate on it. See\n\u003Ca href=\"\u002Fdocs\u002Fcontrol\u002Fgates\">Gates and tool scopes\u003C\u002Fa>.\u003C\u002Fp>\n\u003Cdiv class=\"prose__planned\">\n\u003Cp class=\"prose__flag\">Planned\u003C\u002Fp>\n\u003Cp>What a merge needs will be set on the task's own gate, for any host's merge tool: that the gate is required, that\nthe run writes a summary before it waits, and that approving it ends the run. The OS will not know GitHub's merge\ntool by name.\u003C\u002Fp>\n\u003C\u002Fdiv>\n\u003Ch2 id=\"limits\">Limits\u003C\u002Fh2>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Limit\u003C\u002Fth>\n\u003Cth>Value\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>One command\u003C\u002Ftd>\n\u003Ctd>10 minutes, then it is stopped with everything it started.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>CPUs for one command\u003C\u002Ftd>\n\u003Ctd>4, also given as \u003Ccode>OS_CPU_BUDGET\u003C\u002Fcode>. See \u003Ca href=\"#a-command-plans-for-four-cpus\">A command plans for four CPUs\u003C\u002Fa>.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>A command's output\u003C\u002Ftd>\n\u003Ctd>At most 32,000 characters go back to the member. The exit code always does.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>Setup and teardown\u003C\u002Ftd>\n\u003Ctd>8 steps each, each at most 16 argv entries including the binary.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>Clone depth\u003C\u002Ftd>\n\u003Ctd>The last 50 commits.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>Reading without changing\u003C\u002Ftd>\n\u003Ctd>40 repeated reads (the same file, folder or search again, or an ad hoc command) with no write, edit or commit between them, and a per-run budget of 600 distinct files, folders and searches. The next one past either is refused (\u003Ccode>exploration_budget_exceeded\u003C\u002Fcode>), and the member is told to act on what it has read.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>Search\u003C\u002Ftd>\n\u003Ctd>100 matches by default, 500 at most.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\u003Cp>Running \u003Ccode>mise run …\u003C\u002Fcode>, \u003Ccode>git status\u003C\u002Fcode>, \u003Ccode>git diff\u003C\u002Fcode>, \u003Ccode>git log\u003C\u002Fcode> or \u003Ccode>git commit\u003C\u002Fcode> is progress, not reading: it starts the\ncount of repeated reads again. Nothing resets the per-run budget, but a file, folder or search the run has already\nread never counts against it twice.\u003C\u002Fp>\n\u003Ch2 id=\"troubleshooting\">Troubleshooting\u003C\u002Fh2>\n\u003Cdiv class=\"prose__table\">\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>What you see\u003C\u002Fth>\n\u003Cth>Why\u003C\u002Fth>\n\u003Cth>What to do\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_setup_failed\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>A setup step exited non-zero\u003C\u002Ftd>\n\u003Ctd>Run the same step in a fresh clone of the repository and fix it there, or change \u003Ccode>.os\u002Fworkspace.yml\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_manifest_invalid\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>.os\u002Fworkspace.yml\u003C\u002Fcode> is not in the form above\u003C\u002Ftd>\n\u003Ctd>One step per line, \u003Ccode>- [binary, argument]\u003C\u002Fcode>, under \u003Ccode>setup:\u003C\u002Fcode> or \u003Ccode>teardown:\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_service_unknown:&lt;name&gt;\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>.os\u002Fworkspace.yml\u003C\u002Fcode> asks for a service the OS does not provide\u003C\u002Ftd>\n\u003Ctd>Remove it from \u003Ccode>services:\u003C\u002Fcode>. The one service is \u003Ccode>mysql\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_service_unavailable:mysql\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The deployment has no database server for workspaces, the server could not be reached, or the run's workspace cannot reach the network it is on\u003C\u002Ftd>\n\u003Ctd>Read the reason after the code. The run is blocked, and resumes once the server is there\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_setup_bin:&lt;name&gt;\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>A setup step uses a binary that is not allowed\u003C\u002Ftd>\n\u003Ctd>Run it through \u003Ccode>mise\u003C\u002Fcode>, or a \u003Ccode>package.json\u003C\u002Fcode> script\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>tool_auth_denied:workspace.…\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The host refused the clone or the push\u003C\u002Ftd>\n\u003Ctd>Check that the code host binding's account can write to the repository, and that its token has not expired\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_repo_missing\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>Neither the run's input nor a binding it reaches names a repository\u003C\u002Ftd>\n\u003Ctd>Give the run a \u003Ccode>repo\u003C\u002Fcode>, or set the repository on the code host binding\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_repo_no_clone_address\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The run names a short \u003Ccode>owner\u002Fname\u003C\u002Fcode>, and the nearest code host binding it reaches has no \u003Ccode>git.repoTemplate\u003C\u002Fcode> to expand it\u003C\u002Ftd>\n\u003Ctd>Add \u003Ccode>git.repoTemplate\u003C\u002Fcode> under Custom props on that binding on Tools, or give the run a full \u003Ccode>repo\u003C\u002Fcode> address\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>The clone asks for a password, or the push is refused\u003C\u002Ftd>\n\u003Ctd>No binding the run reaches names the remote's host\u003C\u002Ftd>\n\u003Ctd>Add \u003Ccode>git.host\u003C\u002Fcode> and \u003Ccode>git.username\u003C\u002Fcode> under Custom props on that code host's binding, which holds the token — see \u003Ca href=\"#where-the-repository-lives\">Where the repository lives\u003C\u002Fa>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>default_branch_refused\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The run was working on the default branch\u003C\u002Ftd>\n\u003Ctd>Give the task a \u003Ccode>branch\u003C\u002Fcode> template, or start the run with a \u003Ccode>branch\u003C\u002Fcode> input\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>git_push_missing\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The run finished with work it never pushed\u003C\u002Ftd>\n\u003Ctd>Read the run's last commit and push calls: a refused commit or push says why\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_write_not_allowed\u003C\u002Fcode> or \u003Ccode>workspace_commit_not_allowed\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>A path outside the task's rules\u003C\u002Ftd>\n\u003Ctd>Change the task's \u003Cstrong>Workspace\u003C\u002Fstrong>, or the work\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>A build or test run dies with \u003Ccode>EAGAIN\u003C\u002Fcode>, \u003Ccode>Resource temporarily unavailable\u003C\u002Fcode> or &quot;can't spawn worker thread&quot;\u003C\u002Ftd>\n\u003Ctd>A tool sized its workers from the host's cores, not the command's CPU budget\u003C\u002Ftd>\n\u003Ctd>Give the tool its worker count from \u003Ccode>OS_CPU_BUDGET\u003C\u002Fcode>: see \u003Ca href=\"#a-command-plans-for-four-cpus\">A command plans for four CPUs\u003C\u002Fa>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_resource_exhausted:EAGAIN pids.current=&lt;n&gt; pids.max=&lt;m&gt;\u003C\u002Fcode> on Blockers\u003C\u002Ftd>\n\u003Ctd>A command could not start five times in a row: the run's own commands start more processes and threads than a machine allows\u003C\u002Ftd>\n\u003Ctd>Size your suite's workers from \u003Ccode>OS_CPU_BUDGET\u003C\u002Fcode>, then retry the run. Once or twice in a run's history is a busy machine, and the run went back to the queue by itself\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Ccode>workspace_append_only\u003C\u002Fcode>\u003C\u002Ftd>\n\u003Ctd>The run changed or removed a line in an add-only path\u003C\u002Ftd>\n\u003Ctd>Add lines without changing existing ones, or have the task owner grant a deliberate \u003Ccode>rewriteAllow\u003C\u002Fcode> or \u003Ccode>writeAllow\u003C\u002Fcode> exception\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n",[12,16,19,23,26,29,32,35,38,41,44,47,50,53,56,59,62,65,68,71,74,77,80,83,86,89,92,95],{"id":13,"text":14,"level":15,"planned":8},"what-a-workspace-is","What a workspace is",2,{"id":17,"text":18,"level":15,"planned":8},"where-the-repository-lives","Where the repository lives",{"id":20,"text":21,"level":22,"planned":8},"the-credential-comes-from-the-code-host-tool-you-bind","The credential comes from the code host tool you bind",3,{"id":24,"text":25,"level":22,"planned":8},"credentials-for-the-repositorys-own-commands","Credentials for the repository's own commands",{"id":27,"text":28,"level":15,"planned":8},"the-branch","The branch",{"id":30,"text":31,"level":22,"planned":8},"kept-up-with-the-default-branch","Kept up with the default branch",{"id":33,"text":34,"level":15,"planned":8},"what-the-repository-provides","What the repository provides",{"id":36,"text":37,"level":22,"planned":8},"setup","Setup",{"id":39,"text":40,"level":22,"planned":8},"services-a-database-for-each-run","Services: a database for each run",{"id":42,"text":43,"level":22,"planned":8},"wiring-mise-toml","Wiring mise.toml",{"id":45,"text":46,"level":22,"planned":8},"git-hooks","Git hooks",{"id":48,"text":49,"level":22,"planned":8},"verification-before-merge","Verification before merge",{"id":51,"text":52,"level":22,"planned":8},"commands-for-the-engineering-tasks","Commands for the engineering tasks",{"id":54,"text":55,"level":22,"planned":8},"a-check-has-to-fit-one-command","A check has to fit one command",{"id":57,"text":58,"level":22,"planned":8},"a-command-plans-for-four-cpus","A command plans for four CPUs",{"id":60,"text":61,"level":22,"planned":8},"temporary-files-and-package-stores","Temporary files and package stores",{"id":63,"text":64,"level":22,"planned":8},"commands-for-a-dependency-refresh","Commands for a dependency refresh",{"id":66,"text":67,"level":22,"planned":8},"commands-for-infrastructure","Commands for infrastructure",{"id":69,"text":70,"level":22,"planned":8},"release-copy-os-release","Release copy: os:release",{"id":72,"text":73,"level":22,"planned":8},"public-changelog-os-changelog","Public changelog: os:changelog",{"id":75,"text":76,"level":15,"planned":8},"llm-instruction-for-migrating-an-existing-workspace","LLM Instruction for migrating an existing workspace",{"id":78,"text":79,"level":22,"planned":8},"connect-the-team-after-the-repository-change","Connect the team after the repository change",{"id":81,"text":82,"level":15,"planned":8},"what-a-task-decides","What a task decides",{"id":84,"text":85,"level":15,"planned":8},"commits-and-pushes","Commits and pushes",{"id":87,"text":88,"level":15,"planned":8},"what-is-never-allowed","What is never allowed",{"id":90,"text":91,"level":15,"planned":8},"what-waits-for-you","What waits for you",{"id":93,"text":94,"level":15,"planned":8},"limits","Limits",{"id":96,"text":97,"level":15,"planned":8},"troubleshooting","Troubleshooting",1794928143814]