principle-prove-it-works — Claude Code skill: what it does
Forces the agent to verify task outputs against real artifacts before declaring completion. It prevents models from reporting success just because code compiles.
npx skills add backnotprop/pstack --skill principle-prove-it-works -a claude-codeWhat principle-prove-it-works does
The skill stops the agent from calling work done based on indirect proxies, such as file timestamps, fresh builds, or model self-reports. Indirect checks mislead easily. Instead, the model must run the feature, read the real runtime value, and inspect the actual diff.
Whenever possible, the check must be scripted. A deterministic script produces an artifact that a human reviewer can re-run directly without taking the model's word for it. Committing such verification artifacts makes sense for major migrations or complex ports.
If verification fails, the skill requires checking the observation method before modifying the system under test. This stops the model from rewriting functional code due to an invalid test runner.
What it helps with
- verify outputs against actual artifacts before marking tasks done
- stop agents from closing tickets because the code merely compiles
- generate deterministic scripts to reproduce verification results
- inspect observation scripts before editing functional system code
Example prompts for the agent
- Apply principle-prove-it-works and verify the endpoint response with a real call.
- Write a verification script and show its output before declaring this done.
How to install principle-prove-it-works
1Skills CLI (Vercel)
npx skills add backnotprop/pstack --skill principle-prove-it-works -a claude-code2Manually: put the files in place
~/.claude/skills/principle-prove-it-works/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/principle-prove-it-works/
1Skills CLI (Vercel)
npx skills add backnotprop/pstack --skill principle-prove-it-works -a codex2Manually: put the files in place
.agents/skills/principle-prove-it-works/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add backnotprop/pstack --skill principle-prove-it-works -a cursor2Manually: put the files in place
.cursor/skills/principle-prove-it-works/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install backnotprop/pstack principle-prove-it-works2Manually: put the files in place
.github/skills/principle-prove-it-works/SKILL.md1Skills CLI (Vercel)
npx skills add backnotprop/pstack --skill principle-prove-it-works -a gemini-cli2Platform CLI
gemini skills install https://github.com/backnotprop/pstack --path skills/principle-prove-it-worksManually: put the files in place
.devin/skills/principle-prove-it-works/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add backnotprop/pstack --skill principle-prove-it-works -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about principle-prove-it-works
How do I install principle-prove-it-works in Claude Code?
Run npx skills add backnotprop/pstack --skill principle-prove-it-works -a claude-code. The skill file will be placed in ~/.claude/skills/principle-prove-it-works/SKILL.md.
What is the purpose of principle-prove-it-works?
It prevents agents from assuming work is complete through compiler logs or self-reports. The assistant must directly inspect the live artifact, process state, or output diff.
What should the agent do if verification fails?
The skill states that you must suspect the observation script or environment before suspecting the system. You should not modify working code until you verify the test method itself.
Source text
SKILL.md · MIT
--- name: principle-prove-it-works description: "Apply after completing a task, before declaring done. Verify against the real artifact (run the feature, read the actual value, inspect the diff), not a proxy, self-report, or 'it compiles.'" disable-model-invocation: true --- # Prove It Works Verify every task output by checking the real thing directly. Do not infer from proxies, self-reports, or "it compiles." **Why:** Unverified work has unknown correctness. Indirect verification (file mtimes, output freshness, agent self-reports, cached screenshots) feels cheaper than direct observation. Acting on a wrong inference costs far more than checking the source. Check the real thing, not a proxy: - Check process liveness directly, not indirectly through derived state - Read the actual value, not a cached or derived representation - When verification fails, suspect the observation method before suspecting the system ## Script the check when you can The strongest proof is a deterministic script that re-runs the same comparison, not a one-time eyeball. Write the script, run it, and keep its output as an artifact a reviewer can re-run instead of trusting your word. Keep the artifact visible for the human. Commit it only for large or complex work where the trail has to be auditable later, like a big port or migration (the **show-me-your-work** skill).
© backnotprop · MIT · backnotprop/pstack
