agentleFS
Sign inStart free
raalarcon9705raalarcon9705/gated-dev-agents/skills/verifying-changes/SKILL.md

verifying-changes

Skill0 starsChanged 13 days agoMITView on GitHub

Use when asked to verify, QA, test, validate, review-for-approval or sign off a change, PR, branch, commit, bead, ticket, or another agent's or skill's output, or when about to give a PASS/FAIL, LGTM, "good to go" or ship verdict, in the main thread or as a subagent, especially under time pressure or when tests, screenshots or the author already say it works.

5 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions5 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 5

    1. BuilderIO/agent-native★ 7.6ksame name
    2. PostHog/posthog-foss★ 728same name
    3. Mr-Quin/danmaku-anywhere★ 550same name
    4. teispace/teitunnel★ 2same name
    5. vadimvitvickiy/vv-agent-kit★ 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 repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run it, then give a verdict.

    line 8
  2. Distinctive:

    Read-only: Never fix, revert or stash product code, even if asked.

    line 12
  3. Distinctive:

    UI:: use a real browser (claude-in-chrome or Playwright) on the running app.

    line 16
  4. Distinctive:

    Adversarial: write 3-5 unit tests in a temp file covering boundaries, empty/null, malformed input and error paths.

    line 17

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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

Verifying Changes

Run it, then give a verdict. CI, screenshots and the author's word are claims, not evidence. Builds on superpowers:verification-before-completion.

Rules

  1. Read-only. Never fix, revert or stash product code, even if asked; report it. The only writes allowed are temp adversarial test files, deleted afterwards.
  2. Run it. Build, test, exercise. Code review alone is not QA.
  3. Use the repo's commands: what the caller states in the prompt, else the repo's own build files (package.json, Makefile). Don't depend on CLAUDE.md or workspace docs. Use a test-lock command if the caller gives one. Scoped suites only.
  4. AC table. Each criterion needs evidence (command plus output). One unmet = FAIL. Failing tests or lint = FAIL. An unrelated bug = PASS the item and report the bug separately.
  5. UI: use a real browser (claude-in-chrome or Playwright) on the running app. For a shared component, exercise or trace EVERY mount surface and list them.
  6. Adversarial: write 3-5 unit tests in a temp file covering boundaries, empty/null, malformed input and error paths. A failure is a proven gap: report its code and output. Delete the file, never commit it.
  7. Pre-mortem: dispatch Agent subagent_type: gated-dev:diff-premortem with ONLY the repo path and diff range, never the ticket or intent. Report its 1-3 risks apart from AC. If it is unavailable, do it yourself from the diff only, before re-reading the ticket.
  8. Report. The first word is PASS or FAIL. Then the env, branch and commit tested; the AC table; evidence; adversarial findings; pre-mortem risks.

Can't run something? FAIL (unverified: <what>), never "not signed off yet" or "your call".

Rationalizations

ExcuseReality
"Release cut in 20 min" / "if time runs out, your call"The bar doesn't move. Blocked = FAIL (unverified).
"I'll fix small issues directly, as asked"Verifiers never edit. Report it.
"Stash/revert the fix to see the tests go red"That mutates the tree. Reproduce it in a temp test.
"Author was right the last 10 times"A track record is not evidence for this diff.
"CI green" / "the screenshot shows it"Existing tests and one happy path. Check every AC and every surface.
"Skim the diff" / "run the full suite"A skim is not a pre-mortem. Run scoped suites only.

Red Flags: stop

  • The reply doesn't start with PASS or FAIL
  • About to edit, stash or revert product code
  • An AC lacks output, or only one path/surface was checked
  • The temp test file is still on disk

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.

Check your setup, free