agentleFS
papagrationbiz-sketchpapagrationbiz-sketch/token-scrooge/skills/quality-gate/SKILL.md

quality-gate

Skill0 starsChanged 4 months agoMITView on GitHub

When to load — before declaring any task "done", after a delegated task reports completion, or when writing a completion report. Covers the verification gate, evidence requirements, pre-completion self-review, failure discipline, the planning gate, report format, and the anti-pattern list. Applies to whichever model is currently in the main-thread seat.

46 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use deploy.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • VercelSkill

    deploy

    Deploy applications to Vercel. Use when the user says "deploy", "deploy to Vercel", "push to production", "deploy my app", or "go live".

    Also in 3 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

19 rules. 0 are common, 18 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 36 edited versions we read keep these.

  • 18

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before declaring something "done," actually run the build, test, or execution check yourself, and quote the relevant lines of output.

    line 12
  2. Distinctive:

    If a fix is failing, report it as failing.

    line 14
  3. Very specific:

    Do not write API names, config keys, or command-line options from memory alone — confirm they exist first via grep or --help.

    line 20
  4. Distinctive:

    Before reporting, review the full diff and confirm:

    line 24
  5. Distinctive:

    No out-of-scope changes crept in (no unrequested refactors, reformatting, or import cleanup).

    line 26
  6. Distinctive:

    No leftover debug code, stray prints, or commented-out debris.

    line 27
  7. Distinctive:

    No existing tests were deleted, skipped, or altered just to make them pass.

    line 28
  8. Distinctive:

    No mock/stub substitutions or environment changes made solely to force tests green.

    line 29
  9. Distinctive:

    After two failed attempts with the same approach, change your hypothesis.

    line 33
  10. Distinctive:

    After three failures, stop and report the situation, including the exact error message.

    line 34
  11. Distinctive:

    Never suppress an error with a try/catch that swallows it.

    line 35
  12. Distinctive:

    Never omit outstanding items.

    line 46
  13. Distinctive:

    Don't add a try/catch that swallows an error.

    line 52
  14. Distinctive:

    Don't produce an unrelated diff labeled "cleaned up unused imports." Touch only the files you were asked to touch.

    line 53
  15. Distinctive:

    Don't rewrite a test when it fails.

    line 54
  16. Distinctive:

    Don't add a dependency you weren't asked to add.

    line 55
  17. Distinctive:

    Don't read an entire large file.

    line 56
  18. Distinctive:

    After completing an implementation, an independent cross-model review is required (e.g.

    line 60
  19. Distinctive:

    Do not declare "done" until the Must-fix items from that review are resolved.

    line 61

“Common” is measured: kept by at least 6 in 10 of the 36 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

Quality Gate

Common work-quality rules that apply regardless of which model currently occupies the main thread. The weaker the model in that seat, the more strictly these must be followed. Project-level rules override this skill where they conflict.

1. Verification gate (most important)

  • Before declaring something "done," actually run the build, test, or execution check yourself, and quote the relevant lines of output.
  • Anything you have not run must be labeled "unverified." Never write a completion report based on "it probably works" or "this should work."
  • If a fix is failing, report it as failing. Do not describe a failure as a success.

2. Evidence required

  • Any reference to code must include file:line. Do not describe the contents of a file you have not read.
  • Label speculation explicitly as "guess:" when you have not confirmed it.
  • Do not write API names, config keys, or command-line options from memory alone — confirm they exist first via grep or --help.

3. Self-review immediately before reporting completion

Before reporting, review the full diff and confirm:

  • No out-of-scope changes crept in (no unrequested refactors, reformatting, or import cleanup).
  • No leftover debug code, stray prints, or commented-out debris.
  • No existing tests were deleted, skipped, or altered just to make them pass.
  • No mock/stub substitutions or environment changes made solely to force tests green.

4. Failure discipline

  • After two failed attempts with the same approach, change your hypothesis. Do not keep retrying the same fix.
  • After three failures, stop and report the situation, including the exact error message.
  • Never suppress an error with a try/catch that swallows it. Find the root cause instead.

5. Planning gate

  • Any change touching 3+ files, or involving a design decision, requires a plan presented before starting work.
  • The plan must explicitly state what is in scope to change and what is out of scope.

6. Report format

  • Lead with the conclusion: what you did, what happened, whether it's actually fixed.
  • Required fields: list of changed files / verification results (commands run and their results) / outstanding or unverified items.
  • Never omit outstanding items. Only write "nothing outstanding" if you actually confirmed that.

7. Anti-patterns (living list — append as new failures are observed)

Format: "Don't do X. Do Y instead."

  • Don't add a try/catch that swallows an error. Identify the root cause and fix it instead.
  • Don't produce an unrelated diff labeled "cleaned up unused imports." Touch only the files you were asked to touch.
  • Don't rewrite a test when it fails. Suspect the implementation instead (only touch the test itself once you can show the test itself is buggy).
  • Don't add a dependency you weren't asked to add. Use the standard library or existing dependencies instead.
  • Don't read an entire large file. Read only the symbol and surrounding context you actually need.

8. Additional gate when a weaker model occupies the main-thread seat

  • After completing an implementation, an independent cross-model review is required (e.g. via an external review agent/skill, or a second model).
  • Do not declare "done" until the Must-fix items from that review are resolved.

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.