verification-before-completion
Use when claiming work is complete, tests pass, or fixes work. Required before commits and PRs. Enforces running verification commands and confirming output before success claims. Triggers: 'done', 'fixed', 'passing', 'works now', before committing.
When & Why to Use This Skill
The Verification Before Completion skill implements a rigorous, evidence-based protocol for software development and agentic workflows. It mandates that any claim of task completion, bug fixes, or successful builds must be supported by verifiable command outputs and objective data. By enforcing a 'verification before claim' rule, this skill eliminates guesswork, prevents premature commits, and ensures high-standard code quality by requiring a full 'Red-Green' cycle and comprehensive test suite validation before any status update.
Use Cases
- Pre-Commit & PR Validation: Ensures that the full test suite and linter pass with zero errors before allowing code commits or Pull Request submissions, preventing 'broken' code from entering the repository.
- TDD and Regression Testing: Facilitates a strict Red-Green-Refactor cycle by requiring proof that a regression test failed before the fix and passed after, confirming the fix is effective and targeted.
- Build Integrity Checks: Verifies successful compilation and exit codes across the entire project build process, ensuring that partial successes or linter passes aren't mistaken for a complete, functional build.
- Requirement Compliance Auditing: Performs a systematic, line-by-line verification of project requirements against the actual implementation to ensure all functional criteria are met before marking a task as finished.
- Agent Reliability Guardrail: Acts as a critical safety layer for AI agents, preventing them from reporting success based on 'confidence' or 'intent' rather than actual execution and output analysis.
| name | verification-before-completion |
|---|---|
| description | "Use when claiming work is complete, tests pass, or fixes work. Required before commits and PRs. Enforces running verification commands and confirming output before success claims. Triggers: 'done', 'fixed', 'passing', 'works now', before committing." |
Verification Before Completion
Overview
Claiming work is complete without verification is dishonesty, not efficiency.
Core principle: Evidence before claims, always.
Verification Protocol
Before claiming any status:
- Identify: What command proves this claim?
- Run: Execute the full command (fresh, complete)
- Read: Full output, check exit code, count failures
- Verify: Does output confirm the claim?
- If no: State actual status with evidence
- If yes: State claim with evidence
- Only then: Make the claim
Skip any step = no claim permitted.
Common Failures
| Claim | Requires | Not Sufficient |
|---|---|---|
| Tests pass | Full suite: 0 failures | "My new tests pass", ignoring pre-existing failures |
| Linter clean | Linter output: 0 errors | Partial check |
| Build succeeds | Build: exit 0 | Linter passing |
| Bug fixed | Test symptom: passes | Code changed |
| Regression test | Red-green cycle | Test passes once |
| Agent completed | VCS diff verified | Agent reports success |
| Requirements met | Line-by-line check | Tests passing |
Red Flags
Stop immediately if:
- Using "should", "probably", "seems to"
- Expressing satisfaction before verification ("Great!", "Perfect!", "Done!")
- About to commit/push/PR without verification
- Trusting agent success reports
- Relying on partial verification
- Rationalizing pre-existing test failures as "unrelated"
- Claiming "my tests pass" when full suite has failures
- Any wording implying success without having run verification
Verification Patterns
Tests:
Good: [Run full test suite] [See: 34/34 pass] "All tests in suite pass"
Bad: "My new tests pass" (ignoring 2 pre-existing failures)
Bad: "Should pass now" / "Looks correct"
Regression tests (TDD Red-Green):
Good: Write -> Run (pass) -> Revert fix -> Run (must fail) -> Restore -> Run (pass)
Bad: "I've written a regression test" (without red-green verification)
Build:
Good: [Run build] [See: exit 0] "Build passes"
Bad: "Linter passed" (linter doesn't verify compilation)
Requirements:
Good: Re-read plan -> Create checklist -> Verify each -> Report
Bad: "Tests pass, phase complete"
Agent delegation:
Good: Agent reports success -> Check VCS diff -> Verify -> Report actual state
Bad: Trust agent report
Rationalization Prevention
No shortcuts exist. Counter every excuse:
- "Should work now" -> Run verification
- "I'm confident" -> Confidence is not evidence
- "Just this once" -> No exceptions
- "Linter passed" -> Linter is not compiler
- "Agent said success" -> Verify independently
- "Partial check is enough" -> Partial proves nothing
- "Pre-existing failures are unrelated" -> Full suite must pass, no exceptions
- "My tests pass" -> All tests must pass, not just yours
When To Apply
Always before:
- Any variation of success/completion claims
- Any expression of satisfaction
- Any positive statement about work state
- Committing, PR creation, task completion
- Moving to next task
Rule applies to exact phrases, paraphrases, synonyms, and implications.
Bottom Line
Run the command. Read the output. Then claim the result.
No exceptions.