name: verification-before-completion description: "Use before claiming technical work is complete or successful. Evidence before claims, always." version: 1.0.0 author: Hermes Agent license: MIT platforms: [linux, macos, windows] metadata: hermes: tags: [verification, testing, validation, quality-assurance, completion-gate] related_skills: [systematic-debugging, test-driven-development, writing-plans]

Verification Before Completion

Overview

Claiming work is complete without fresh verification evidence is unacceptable.

Core principle: Evidence before claims, always.

The Iron Law

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE

If you haven't run the verification command in the current turn, you cannot claim it passes.

The Gate Function

Before claiming any task is done, expressing satisfaction ("Done!", "Works!", "Fixed!"), or committing:

1. IDENTIFY: What exact command, test, or output proves this claim?
2. RUN: Execute the FULL verification command (fresh, complete, no stale cached results).
3. INSPECT: Read the full stdout/stderr, verify exit status (code 0), check log output and file contents.
4. VERIFY: Does the actual evidence prove the requirements are met?
5. If NO: Report the actual failure or gap with exact error output.
6. If YES: State the completion claim WITH the supporting evidence.
7. ONLY THEN: Mark the task as completed.

Verification Checklist

• [ ] Requirements: Verified each user requirement individually against code/behavior.
• [ ] Direct Execution: Ran the code, service, or CLI command directly.
• [ ] Automated Tests: Ran relevant unit tests and confirmed 0 failures.
• [ ] Regressions: Checked for obvious unintended side-effects in related modules.
• [ ] Status Classification: Clearly distinguish between:
• VERIFIED: Proven by direct fresh command output.
• LIKELY: High confidence but indirect evidence.
• UNVERIFIED: Assumed or untested.
• FAILED: Did not produce expected output.