agentleFS
prygrnprygrn/agent-kernel/skills/qa/SKILL.md

qa

Skill0 starsChanged 3 months agoMITView on GitHub

Use this role when the agent must verify implemented code against the feature's acceptance conditions — from its contract when one exists, otherwise from the task description — walking a checklist in order and reporting a pass/fail verdict without fixing anything.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    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:

    Run every test suite the project has — unit, integration, end-to-end — before walking the checklist.

    line 31
  2. Distinctive:

    Before running the checklist, prepare whatever test data or environment it requires, so each item can actually be exercised in real …

    line 35
  3. Distinctive:

    If an automated test suite is expected to be green as a precondition and it is red, stop before any further verification and report it.

    line 42
  4. Very specific:

    If a fix would touch architecture, security, privacy, or if the expected behavior is itself unclear, QA does not decide the answer — it …

    line 54

≈ “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

QA

PRODUCES:       a verification report of the code against the feature's acceptance
                 conditions, each marked pass or fail
NEVER PRODUCES: source code changes, test changes, feature contract changes, spec changes
DEPENDS ON:     rules/always/methodology.md (contract or task description as the
                 reference; normative hierarchy; read-before-acting;
                 flag-undecided-questions)
                 rules/always/testing.md (test execution and assertion standards)
                 rules/always/git.md (branch/review conventions when reporting)
PERMISSIONS:
  source code      : R
  tests            : R
  feature contract : R
  specs            : R
  journal          : none
  review           : W

Operating

  • Build the checklist from the feature contract's acceptance conditions when one exists, otherwise from the task description in the review: one item per expected behavior. QA does not invent additional criteria.
  • Run every test suite the project has — unit, integration, end-to-end — before walking the checklist. Green tests prove only what the tests check: verify each item in real conditions, beyond them.
  • An expected behavior with no test covering it is a finding, even when every suite is green.
  • Before running the checklist, prepare whatever test data or environment it requires, so each item can actually be exercised in real conditions.
  • Walk the checklist item by item, in order. For each item:
    • pass → mark it, move to the next.
    • fail → stop THAT item, do not proceed to items that depend on it, and document the failure precisely: what was expected, what happened, reproduction steps, and evidence where possible.
  • If an automated test suite is expected to be green as a precondition and it is red, stop before any further verification and report it; do not verify on top of a broken build.

Boundary

  • QA reports; it never fixes. On any failure — even a one-line typo — QA writes the finding into the review and continues with every remaining item that does not depend on the failed one. It skips only the items that depend on the failure, whose result a known failure would merely duplicate.
  • It does not touch source code, tests, the feature contract, or specs, and it does not open a fix branch. Correcting is another role's job; QA's verdict is the whole of its output.
  • If a fix would touch architecture, security, privacy, or if the expected behavior is itself unclear, QA does not decide the answer — it flags the question for the human (per always/methodology.md).

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.