agentleFS
anotherbenanotherben/claude-enterprise-skills/skills/enterprise-verify/SKILL.md

enterprise-verify

Skill1 starsChanged 8 months agoMITView on GitHub

Evidence-based verification before any completion claim. No 'should work' or 'probably fine' — paste fresh test output or don't claim done. 7-check verification sequence. Use before committing or claiming work is complete.

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are 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.

  • 2

    Very specific

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

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

    If overall is FAIL: stop, fix the failing checks, re-run verify.sh from scratch.

    line 55
  2. Distinctive:

    When a check fails:

    line 116
  3. Distinctive:

    If tests fail after a fix: you introduced a regression — revert and try again

    line 119
  4. Distinctive:

    If you catch yourself reaching for these phrases, you haven't run verification yet:

    line 125
  5. Very specific:

    Test suite takes too long?: Run verify.sh once with full suite.

    line 139

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Enterprise Verify — Evidence-Based Completion Verification

Why Evidence Matters

Without verification evidence, completion claims are beliefs — and beliefs ship bugs. "The code looks correct" means nothing when an import path is wrong, a test is silently skipped, or a debug console.log made it into the diff. Every check category below has caused a production incident at least once.

The distinction is simple: evidence is command output you can paste. Belief is anything else. This skill deals only in evidence.


Step 1: Run verify.sh

The verification script runs all 7 checks mechanically and produces structured JSON evidence. Run it — don't run the checks manually.

bash .claude/skills/enterprise-verify/scripts/verify.sh \
  --base dev \
  --contract <path-to-contract> \
  --output .claude/enterprise-state/<slug>-verification-evidence.json

Flags:

  • --base <branch> — base branch for diffs (default: dev)
  • --contract <path> — contract file for postcondition extraction (optional)
  • --skip-build — skip frontend build check (backend-only changes)
  • --output <path> — where to write evidence JSON
  • --project <path> — project root (auto-detected from cwd)

Step 2: Read the JSON Evidence

The script outputs a JSON file with this structure:

{
  "checks": {
    "test_suite":         { "result": "PASS|FAIL", "passed": N, "failed": N },
    "postcondition_trace": { "result": "MANUAL", "postconditions": "...", "test_names": "..." },
    "regression":          { "result": "PASS|FAIL", "new_failures": N },
    "build":               { "result": "PASS|FAIL|SKIP" },
    "diff":                { "result": "PASS", "files": [...] },
    "imports":             { "result": "PASS|FAIL", "unresolved": [...] },
    "debug_artifacts":     { "result": "PASS|FAIL", "findings": [...] }
  },
  "overall": "PASS|FAIL"
}

If overall is FAIL: stop, fix the failing checks, re-run verify.sh from scratch.


Step 3: Complete the Two Manual Checks

The script automates 5 of 7 checks fully and collects data for the other 2. You must complete:

Postcondition Trace (Check 2)

Map each postcondition from the contract to a specific test name from the test_names field. Use the EXACT test description from the runner output — not a paraphrase. If a postcondition has no matching test: FAIL.

Diff Classification (Check 5)

The script lists changed files. Classify each as:

  • REQUIRED — in the contract/plan
  • ENABLING — needed to support a required change
  • DRIFT — not related to this task → git checkout -- path/to/file

Step 4: Write the Verification Report

===============================================
         ENTERPRISE VERIFICATION REPORT
===============================================

Task: [what was done]
Date: [YYYY-MM-DD]
Branch: [branch name]
Evidence: [path to JSON file]

## Automated Checks (from verify.sh)
  Check 1 — Test Suite:          [PASS — N passed, 0 failed]
  Check 3 — Regression Check:    [PASS — no regressions]
  Check 4 — Build Verification:  [PASS / SKIPPED (backend only)]
  Check 6 — Import Resolution:   [PASS — all imports resolve]
  Check 7 — Debug Artifacts:     [PASS — none found]

## Manual Checks
  Check 2 — Postcondition Trace: [PASS — N/N verified]
  Check 5 — Final Diff:          [PASS — N files, 0 drift]

  ────────────────────────────
  OVERALL: [PASS / FAIL]

When to Run

TriggerAction
About to say "it's done"Run verify.sh first
About to commitRun verify.sh first
About to create a PRRun verify.sh first
"One more thing" was fixedRe-run verify.sh from scratch

Failure Recovery

When a check fails:

  1. Fix the issue
  2. Re-run verify.sh entirely (not just the failed check)
  3. If tests fail after a fix: you introduced a regression — revert and try again

Language That Signals Missing Evidence

If you catch yourself reaching for these phrases, you haven't run verification yet:

  • "should work" / "probably fine" → run verify.sh, read the JSON
  • "looks good" / "looks correct" → run verify.sh, report evidence
  • "I'm confident" / "I believe" → confidence isn't evidence

Edge Cases

No formal contract? Derive postconditions from the task description. Run verify.sh without --contract — it still runs all other checks.

Only one file changed? Still run verify.sh. A one-line change can break tests or leave debug artifacts.

Test suite takes too long? Run verify.sh once with full suite. Don't substitute partial runs.

See references/verification-checks.md for the detailed command reference behind each check.

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.