Agent skill
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.
Install this agent skill to your Project
npx add-skill https://github.com/Delphine-L/claude_global/tree/main/skills/claude-meta/verification-before-completion
SKILL.md
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.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
Delphine-L/claude_global
conda-recipe
Expert in building and testing conda/bioconda recipes, including recipe creation, linting, dependency management, and debugging common build errors
Delphine-L/claude_global
bioinformatics-visualization
Publication-quality bioinformatics figures - phylogenetic trees, genome browsers, iTOL datasets, and data presentation
Delphine-L/claude_global
phylogenetics
Phylogenetic tree analysis, visualization, annotation management, and iTOL troubleshooting
Delphine-L/claude_global
bioinformatics-fundamentals
Core bioinformatics concepts including SAM/BAM format, AGP genome assembly format, sequencing technologies (Hi-C, HiFi, Illumina), quality metrics, and common data processing patterns. Essential for debugging alignment, filtering, pairing issues, and AGP coordinate validation.
Delphine-L/claude_global
project-sharing
Prepare organized packages of project files for sharing at different levels - from summary PDFs to fully reproducible archives. Creates copies with cleaned notebooks, documentation, and appropriate file selection. After creating sharing package, all work continues in the main project directory.
Delphine-L/claude_global
hackmd
HackMD collaborative markdown - slide presentations, embedded SVG diagrams, and real-time editing best practices
Didn't find tool you were looking for?