OpenParable
Skip to content
← Back to skills

Self Verification

Testing & Quality
self-verification

Self Verification is an Agent Skill for testing & quality published in the A-EVO-Lab/a-evolve repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Verification patterns to confirm task completion before submitting. Read this before calling submit().

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/A-EVO-Lab/a-evolve/blob/HEAD/artifacts/tb2_clawcode_opus46/skills/self-verification/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/self-verification, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Self-Verification Checklist

Before calling submit(), systematically verify your solution:

1. Re-read Requirements

  • Open and re-read the original task description
  • List each specific requirement (files to create, formats, thresholds, etc.)

2. Verify Each Requirement

For each requirement, run a concrete check:

  • File creation: ls -la <path> and head <path>
  • Code changes: Run the modified code and check output
  • Build tasks: Run the built binary with a test input
  • Server/web tasks: curl or wget to test endpoints — verify response content, not just that the port is open. Checking ports with ss or netstat is NOT enough
  • Data tasks: Check output format, row counts, value ranges
  • Config tasks: If necessary, restart the service and verify it actually works end-to-end

3. Check Your Assumptions

  • If you chose between multiple approaches (e.g., normalization methods, algorithms), verify your choice matches what the task/test expects — don't assume "standard" is correct
  • If you installed or modified system packages, run apt --fix-broken install and verify the package manager still works
  • If your solution works on the provided example, consider whether it generalizes to different inputs (larger, different distributions, edge cases)

4. Common Pitfalls

  • Git tasks: check ALL branches and history, not just HEAD
  • Security tasks: verify secrets are removed from git history too
  • Build tasks: verify no unwanted dependencies (ldd, nm)
  • LaTeX: recompile and check ALL warnings, not just the first