runtime-watch
Skill11 starsChanged 49 days agoMITView on GitHub
DorkOS runtime release watch, called from @dorkos-qa's heartbeat: compares the pinned Claude Agent SDK, Codex SDK and CLI, OpenCode SDK and Doe's Pi packages with npm, reads what changed, says whether it matters to DorkOS, and hands real drift to DorkOS Code for /app:runtime-upgrade. Reads and reports only. Use only when a heartbeat checklist or a person asks for the runtime watch.
In this repository44 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .claude/CLAUDE.mdCLAUDE.md
- auditing-uiSkill
- browser-testingSkill
- capturing-product-mediaSkill
- ci-local-exportSkill
- ci-pulseSkill
- ci-quarantineSkill
- ci-statusSkill
- creating-pull-requestsSkill
- debugging-systematicallySkill
- debugging-test-failuresSkill
- debugging-typescript-errorsSkill
- deps-weeklySkill
- designing-frontendSkill
- opensrcSkill
- organizing-fsd-architectureSkill
- stewarding-ci-pipelineSkill
- syncing-agent-skillsSkill
- verification-before-completionSkill
- vision-drift-checkSkill
- visual-companionSkill
- writing-adrsSkill
- writing-app-copySkill
- writing-blog-postsSkill
- writing-changelogsSkill
- writing-developer-guidesSkill
- writing-for-humansSkill
- writing-to-usersSkill
- adding-config-fieldsSkill
- clarifying-requirementsSkill
- maintaining-dependenciesSkill
- maintaining-dev-playgroundSkill
- managing-specsSkill
- marketplace-devSkill
- orchestrating-parallel-workSkill
- reading-session-transcriptsSkill
- receiving-code-reviewSkill
- requesting-code-reviewSkill
- styling-with-tailwind-shadcnSkill
- test-driven-developmentSkill
- upgrading-runtime-dependenciesSkill
- working-in-worktreesSkill
- dorkbot-triageSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Stop early when nothing moved: If every package is none, the report is one line and the run is over.
- Very specific: line 24
Check it is not already in hand: Look for an open PR whose title names the package (gh pr list --search "<short_name> in:title") and an …
- Very specific: line 34
When nothing moved: Runtimes (<date>): all current.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
runtime-watch
A cheap, read-only check: a few npm view calls and, only when something is new, one read of the release notes. It changes no file, opens no PR and needs no node_modules, so it is safe to run every day. The bump itself is DorkOS Code's job, through /app:runtime-upgrade (.claude/commands/app/runtime-upgrade.md) and the upgrading-runtime-dependencies skill.
What it watches
The packages in .claude/config/runtime-deps.json (@anthropic-ai/claude-agent-sdk, @openai/codex-sdk, @openai/codex, @opencode-ai/sdk) and the Pi packages Doe uses (@earendil-works/pi-agent-core, pi-ai, pi-mcp in packages/doe/package.json).
Run
- Read the pins from
origin/main, not a working tree.git fetch origin, pinBASE=$(git rev-parse origin/main), then read each version withgit show "${BASE}:apps/server/package.json"(the Pi packages:"${BASE}:packages/doe/package.json"). Write${BASE}with braces: in zsh,$BASE:ais a path modifier. Strip a leading~or^. - Read the latest. For each package,
npm view <pkg> versionandnpm view <pkg> time --jsonfor the publish date. Ignore pre-release tags. - Classify the drift:
none,patch,minorormajor. In a0.xline, a minor bump (0.154 → 0.163) counts as a minor with breaking risk. For the Claude Agent SDK, every0.3.xpatch ships features and fixes, so read its notes whatever the size. - Stop early when nothing moved. If every package is
none, the report is one line and the run is over. - For each package that moved, read what changed between the pin and the latest, from the sources listed for it in
runtime-deps.json(npm, GitHub releases,CHANGELOG.md). Sort it into breaking, deprecated, new and fixed. Then decide whether it matters to DorkOS:- Matters now: a breaking change or deprecation on a surface in that package's
sdk_surface_map; a fix for something anupgrade_notesentry works around (for example, upstream #454 on the Claude SDK); a security fix; or a new model or model alias. - Worth adopting: a new feature a DorkOS user would notice. Name it in one plain line.
- Routine: everything else.
- Matters now: a breaking change or deprecation on a surface in that package's
- Check it is not already in hand. Look for an open PR whose title names the package (
gh pr list --search "<short_name> in:title") and an open Linear ticket for the same bump. If one exists, report it as "in hand" with its title, and do not hand it off again. - Hand off real drift. A moved package that is not in hand goes to DorkOS Code (
@dorkos-2) as one message: package,pin → latest, the "matters now" lines, and "run/app:runtime-upgrade <pkg>within 2 days". Each runtime bump is its own PR behind its adapter directory, with the runtime conformance suite passing. Every runtime bump opens with theholdlabel and stays unarmed until @security-auditor OKs it in #dorkos, even a plain version bump (Security rule, 2026-10-20). Say so in the hand-off. - New frontier-model support (a new model alias, a new model family) also gets one line to @dorkos-marketing, but only after the bump that carries it has merged.
Report
One line per run, in #dorkos or to whoever called it:
Runtimes (<date>): claude-agent-sdk 0.3.280 → 0.3.304 (matters: <why>), codex-sdk 0.154 → 0.163 (in hand: <PR title>), opencode 1.18 → 1.19 (routine), Pi current.
When nothing moved: Runtimes (<date>): all current.
Never
- Edit a manifest, the lockfile or
runtime-deps.json, open a PR, or file a ticket for a bump (DorkOS Code files those). - Install anything, run an eval, or spend money.
- Hand the same release off twice: step 6 is what keeps a daily run quiet.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

