Agent Skills: Verification Before Completion

Enforces evidence-based completion claims. Use before claiming work is done, tests pass, or bugs are fixed. Requires running verification commands and confirming output before any success claims.

UncategorizedID: delphine-l/claude_global/verification-before-completion

Install this agent skill to your local

pnpm dlx add-skill https://github.com/Delphine-L/claude_global/tree/HEAD/skills/claude-meta/verification-before-completion

Skill Files

Browse the full folder contents for verification-before-completion.

Download Skill

Loading file tree…

skills/claude-meta/verification-before-completion/SKILL.md

Skill Metadata

Name
verification-before-completion
Description
Enforces evidence-based completion claims. Use before claiming work is done, tests pass, or bugs are fixed. Requires running verification commands and confirming output before any success claims.

Verification Before Completion

Overview

Core principle: Evidence before claims, always.

The Gate

BEFORE claiming any status:

1. IDENTIFY: What command proves this claim?
2. RUN: Execute the command (fresh, complete)
3. READ: Full output, check exit code
4. VERIFY: Does output confirm the claim?
   - If NO: State actual status with evidence
   - If YES: State claim WITH evidence
5. ONLY THEN: Make the claim

Common Failures

| Claim | Requires | Not Sufficient | |-------|----------|----------------| | Tests pass | Test command output: 0 failures | Previous run, "should pass" | | Linter clean | Linter output: 0 errors | Partial check | | Build succeeds | Build command: exit 0 | "Looks good" | | Bug fixed | Original symptom: gone | Code changed, assumed fixed | | Pipeline complete | Output files exist and are valid | Tool reported success | | Notebook runs | All cells executed without error | Some cells ran |

Red Flags — STOP

If you catch yourself:

  • Using "should", "probably", "seems to"
  • Expressing satisfaction before verification ("Great!", "Done!")
  • About to commit without running tests
  • Relying on partial verification
  • Trusting a tool's success report without checking output

Rationalizations

| Excuse | Reality | |--------|---------| | "Should work now" | RUN the verification | | "I'm confident" | Confidence != evidence | | "Just this once" | No exceptions | | "Tool said success" | Verify independently | | "Partial check is enough" | Partial proves nothing |

Key Patterns

Tests:

CORRECT: [Run test] → [See: 34/34 pass] → "All tests pass"
WRONG:   "Should pass now" / "Looks correct"

Pipeline output:

CORRECT: [Check output file exists] → [Verify size/format] → "Pipeline produced valid output"
WRONG:   "Galaxy shows green" (without checking actual data)

Notebook:

CORRECT: [Restart kernel] → [Run All] → [Check no errors] → "Notebook runs cleanly"
WRONG:   "I updated the cell" (without re-running)

The Bottom Line

Run the command. Read the output. THEN claim the result.

Attribution

Adapted from obra/superpowers verification-before-completion skill.