
verification-before-completion
by mway
SKILL.md
name: verification-before-completion description: Use when about to claim work is complete, fixed, or passing, or before committing/creating PRs; requires fresh verification evidence before any success claim. allowed-tools: ["shell", "read_file", "update_plan"] metadata: short-description: Evidence gate before completion claims
Verification Before Completion
Read these references:
~/.config/agent/core/behavior.md- Evidence-based reasoning~/.config/agent/core/methodology.md- Verify phase discipline~/.config/agent/core/communication.md- Accurate status reporting
Overview
Claiming work is complete without verification is unacceptable. Evidence comes first.
Core principle: Evidence before claims, always.
Mode-Aware Verification
Verification requirements apply to ALL modes, but execution differs:
Normal mode:
- Confirm before running verification commands
- Show user the commands that will be run
- Execute after confirmation
Autopilot mode:
- Auto-run verification commands (Tier 3 operations)
- Report progress: "Running tests..."
- Show results automatically
Full-auto mode:
- Auto-run verification commands
- Log to audit trail
- Only pause if verification fails repeatedly
Verification failure handling:
- Normal: Report failure, ask for guidance
- Autopilot: Attempt one auto-fix, then ask for guidance
- Full-auto: Attempt up to 3 auto-fixes, then stop and report (see
~/.config/agent/domain/coding/guardrails.mdrecovery procedures)
The Iron Law
NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
If you haven't run the verification command in this message, you cannot claim it passes.
The Gate Function
BEFORE claiming any status or expressing satisfaction:
- Identify the command that proves the claim
- Run the full command (fresh, complete)
- Read the full output and exit code
- Verify the output supports the claim
- If NO: state actual status with evidence
- If YES: state the claim with evidence
- Only then make the claim
Skipping any step is a failure to verify.
Common Failures
| Claim | Requires | Not Sufficient |
|---|---|---|
| Tests pass | Test output shows 0 failures | Previous run, "should pass" |
| Linter clean | Linter output shows 0 errors | Partial check |
| Build succeeds | Build command exit 0 | Linter passing |
| Bug fixed | Test of original symptom passes | Code changed |
| Requirements met | Checklist verified | "Looks done" |
Red Flags — STOP
- Using "should", "probably", "seems"
- Expressing satisfaction before verification
- About to commit/PR without verification
- Trusting agent success reports without evidence
- Relying on partial checks
Examples
Tests:
✅ Run tests → see 0 failures → "All tests pass"
❌ "Should pass now"
Build:
✅ Run build → exit 0 → "Build passes"
❌ "Linter passed"
Requirements:
✅ Re-read plan → verify checklist → "Requirements met"
❌ "Seems complete"
If verification fails or behavior is unclear, switch to systematic-debugging.
Common Verification Commands (Quick Reference)
| Language/Tooling | Command |
|---|---|
| Go | go test ./... |
| Python | pytest |
| Node | npm test |
| Rust | cargo test |
When to Apply
Always before:
- Any completion claim
- Any positive status statement
- Committing, PR creation, or moving to next task
Arguments
Target: ${ARGUMENTS}
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon