verify-gate
Skill9 starsChanged 5 days agoApache-2.0View on GitHub
Run and interpret Fishhawk verification safely. Use before push/PR, or when asked to run verify, tests or the CI gate.
7 other projects use a version of this.
Adopted by
Most recently taken up by
- Ces107/agent-forgetheir own version, first seen Oct 25, 2026
- Jaxsbr/sabstheir own version, first seen Oct 20, 2026
- feelthefusion/skill-starter-kittheir own version, first seen Oct 13, 2026
- limoaCatherine/game-studio-harnesstheir own version, first seen Sep 20, 2026
- svy04/ballasttheir own version, first seen Aug 23, 2026
- piprail/piprailtheir own version, first seen Jun 10, 2026
Other versions7 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 7
- pskoett/pskoett-ai-skills★ 327same name
- svy04/ballast★ 73same name
- piprail/piprail★ 9same name
- limoaCatherine/game-studio-harness★ 4same name
- Ces107/agent-forge★ 0same name
- feelthefusion/skill-starter-kit★ 0same name
- Jaxsbr/sabs★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- deploy-localSkill
- dev-cleanSkill
- diagnose-redSkill
- frontend-checkSkill
- new-go-moduleSkill
- record-adrSkill
- spec-changeSkill
- sync-mainSkill
- teardown-localSkill
The file
Files
Run the verify gate
scripts/verify-change is the repository committed-tree gate: path-aware
frontend checks followed by scripts/test verify. Offline dependencies, exit 75,
stage order and coverage contracts are owned by scripts/README.md and routed
by the verification guide. Lint and static
checks run before the test loop; scoped and no-tests runs do not establish a full
pass. Avoid maintaining a second copy of the gate's leg list here.
1. Preflight — do all of these first
pgrep -fl '[f]ishhawk-runner .*--run-id' || echo no-live-runner
uptime; sysctl -n hw.ncpu
go env GOVERSION
scripts/check-gate-image --host
docker info >/dev/null 2>&1 && echo docker-ok || echo docker-DOWN
git status --short- Live runner. It may be holding the per-repository verify lock. A shell
verifythat meets a live holder refuses immediately. That refusal is final for that invocation: do NOT retry or poll it. Narrow toscripts/test single -run TestX ./path/to/pkg/...instead, or wait for the run to finish. - Load average above ~4× the core count. The host is starved, and unrelated tests will flake. Use the
diagnose-redskill (orphaned agent processes) before running anything. - Toolchain mismatch. The host checker compares the exact lint version, its build Go and standalone gofmt with the checked-in gate-image pins. Treat
host_toolchain_mismatchas environmental, never a reason to edit code. Put the matching tools first on PATH or use--in-gate-image. Read the current Go version fromdeploy/gate-image/Dockerfile; prefer a per-command toolchain/PATH change. Changing persistentgo env -wsettings needs authorization and does not switch standalone gofmt. - Postgres prerequisite. Host tests use Docker/testcontainers. A runner gate container instead uses its supplied
FISHHAWK_TEST_PG_URLand has no Docker daemon; a missing or failing service there is a failure, never a skip. - Dirty tree. That's allowed: patch coverage diffs merge-base → work tree, so uncommitted edits are gated too.
2. Run
| Situation | Command |
| Before a PR / final check (outside a run) | scripts/verify-change |
Inside a Fishhawk run (scripts/is-run-agent prints run-agent, i.e. FISHHAWK_RUN_AGENT is set) whose stage configures a verify gate | Focused scripts/test single plus a scoped scripts/test verify --packages … only. The runner's committed-tree gate runs the full verify on the passing SHA; don't spend the agent budget repeating it (#5148). The marker only selects this workflow and never replaces the gate: with no configured gate, run the full verify yourself |
| Iterating on known packages | scripts/test verify --packages backend/internal/foo,runner/internal/bar (repo-relative). It skips patch coverage, so finish with a full run, which inside a run is the runner's gate |
| One test | scripts/test single -run TestName ./backend/internal/foo/ |
| Lint only | scripts/test lint |
A full verify takes many minutes. Run it in the background (or with your shell tool's longest timeout) and wait for it to exit. Don't start a second concurrent verify.
3. Read the result
| Failure | Meaning / fix |
| golangci-lint / gofmt | Fix the code; the matching pinned gofmt -w / goimports -w on the named files |
Schema sync drift | A docs/spec/ canonical file changed without mirroring it: run scripts/sync-schemas and commit the mirrors |
| Doc-line budget | A docs/ARCHITECTURE.md line is over 1000 chars; move the prose to the package README |
check-adr | Record/index mismatch: scripts/check-adr --write-index (see the record-adr skill) |
Site gates (check-site-voice, check-site-ia, site reference drift) | Banned §5 term on the site, a sidebar/page mismatch, or a stale generated region (scripts/gen-site-reference) |
TestKnownCategoriesCoversEmittedCategories | A new audit category: register it in backend/internal/audit/categories.go |
TestCrossModuleWireParity | A cross-module wire struct drifted: see backend/internal/wirecontract/README.md |
| Patch coverage below 85% | Add tests that exercise the changed lines. Never set FISHHAWK_SKIP_PATCH_COVERAGE to get a pass |
| A test unrelated to the diff | Likely environmental (host load, toolchain, Docker): use the diagnose-red skill before touching code |
host_toolchain_mismatch | Fix the tool environment or use the gate image; do not change code |
| Refused: verify lock held | See preflight. Don't retry |
A skipped harness (zsh or helm absent) is a printed skip, not a failure. Report it as an uncovered area.
4. Report
State pass/fail and the failed gate leg. If the test loop ran, report every failing module and its failing tests/build errors from the aggregate summary. Include the relevant diagnostics and every skip; distinguish scoped/no-tests evidence from a full pass.
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.

