agentleFS
hurttlockerhurttlocker/o8/.agents/skills/o8-checks/SKILL.md

o8-checks

Skill59 starsChanged 43 days agoMITView on GitHub

Choose or run verification for code, process seams, authorization, persistence, or documentation changes.

  • Installs packages

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 0 are distinctive, and 4 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.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every code change runs npm run typecheck:fast and the hermetic npm test completion gate.

    line 8
  2. Very specific:

    Use npm run typecheck:tsc for the slower cross-check.

    line 10
  3. Very specific:

    Run npm run rule-check -- --base=<ref> for changed TypeScript or TSX and ESLint only on touched files.

    line 18
  4. Very specific:

    Run npm run verify:ci before pushing to execute the local CI checks.

    line 23

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Verification

  • Every code change runs npm run typecheck:fast and the hermetic npm test completion gate. typecheck:fast falls back to tsc 5.8 on Intel Macs and Windows. Use npm run typecheck:tsc for the slower cross-check. Run the smallest relevant tests while iterating, and use integration or Rust tests when the changed boundary requires them. Docs-only changes validate links and the affected documentation checks.
  • New cross-process seams, prompt-taught arguments, persistence paths, and principal or authorization changes need a test through the real entry point and persisted state. Helper-only tests are not reachability proof.
  • Run npm run rule-check -- --base=<ref> for changed TypeScript or TSX and ESLint only on touched files. Read .github/workflows/ci.yml before describing current CI.

Local CI parity

Run npm run verify:ci before pushing to execute the local CI checks. Use Node 22, the npm version in packageManager, installed dependencies, and bash and cargo-audit on PATH (install with cargo install cargo-audit --locked). Public dependency audits need network access but no account credentials.

The command reads .github/workflows/ci.yml for Type Check, Lint, Test Classification, all Hermetic Unit Tests shards, Server Bundle Size, Integration Tests (changed files), Governance Smoke, and Security Audit. Fetch origin/main before running. Changed integration files use CI's merge-base selector, including staged, unstaged and untracked classified tests locally, and the same integration configuration and baseline as CI. Install ripgrep and the system secure-shell tools for resource-owning tests. Linux CI restores prerequisite .deb archives for Rust and resource tests using the package definitions, installer version, architecture and exact runner image version. Every run refreshes signed APT indexes and verifies cached bytes against the exact package/version/architecture SHA256 before normal APT installation. Missing, stale or damaged payloads are downloaded by APT. Only main saves shared caches; authenticated archive revisions keep later security updates reusable. This privileged runner setup is explicitly hosted in the coverage guard; local verification uses the installed system prerequisites. The package-cache proof workflow transfers a short-lived archive directory between fresh containers on one hosted runner to measure miss and authenticated hit behavior against an identical base image, without allowing a PR to save the shared cache. Browser tests declare Chromium through their runtime import, or a first-line // @o8-test-browser marker for an indirect launcher. Classification generates browser: true; CI and local verification select the prerequisite from that same metadata. Adding a browser test does not require editing the workflow or verifier. Both coverage and classification guards trace runtime imports, called exports and Node script launches. An indirect browser launcher requires the first-line marker; the refusal names its test and the marker to add. A file-wide mocked launch does not require a browser, while restoring the real launch restores the prerequisite. The coverage guard also rejects browser flags that disagree with the source. It sets CI=true, uses the CI test configuration and environment, gives type checking an 8192 MiB heap, and runs the smoke in a fresh temporary data directory with its CI timeout. Each check prints PASS or FAIL under its CI name; any failed command makes the final exit status non-zero. Audit failures also block CI. Independent checks run concurrently; governance waits for Type Check. Vitest commands share one local execution queue so their module caches and worker pools do not collide.

Dependency changes use the same credential-free lockfile audit in CI and locally: added package versions fail at low severity, while unchanged baseline findings remain the responsibility of the full audits. Nested package and native lockfiles are included. The o8 next Native Tests (Linux) check also runs on Linux x64, using CI's locked Cargo build for the HTTP fixture daemon and all four headless suite shards. CI runs the shards on separate jobs; local verification runs them serially against one Cargo target cache, retaining every discovered test exactly once. Each native shard shares CI's combined daemon-build/test deadline. Native fixed-rate terminal probes run without sibling native tests and share the local Vitest execution queue, retaining their original byte and time bounds. Local queue waiting pauses the native job clock, matching CI's independent runners. Native build/cache outputs are excluded from lint and type checking; source fixtures remain checked. Install the system prerequisites listed in scripts/ci/install-native-linux-deps.sh and the stable Rust toolchain. Its selector compares against origin/main and includes committed, staged, unstaged and untracked native inputs. The selector derives fixture inputs from native source file references, transitive host imports and discovered fixture directories; deleting an input also selects the suite. Headless CI installs only missing compile prerequisites, leaving capture tools and drivers to graphical jobs. A TS-only change outside that dependency closure prints SKIP and runs no native build. Other platforms print HOSTED with the Linux x64 requirement. Hosted reporting, other native target builds and weekly integration lanes remain hosted checks. The CI coverage guard rejects an unclassified job or an unsupported action in a local check. Update the local implementation or its explicit hosted classification whenever CI changes; unknown checks never silently pass.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.