Check whether your AI coding agent really did what it says: compare its pull request claims ('all tests pass') against what actually changed and get a plain-English receipt: AI agent prompt
AI coding agents often report that everything works, and sometimes they get there by deleting or skipping the tests. On Oct 24, 2026 a developer shared proofslip, a free, open source GitHub Action that checks what an AI agent claimed in its pull request against what GitHub actually recorded, then writes a plain-English receipt anyone can read. The prompt below does the same check by hand for any pull request or agent summary, so non-experts can spot claims that the actual changes don't back up before merging.
The prompt
I'm going to paste what my AI coding agent said it did, and what actually changed. Act as a careful reviewer and tell me in plain English whether its claims hold up. Agent's summary: [paste the PR description or the agent's final message]. Changed files and diff: [paste the diff or file list with added/removed lines]. Test or check results: [paste CI output, or 'none']. 1) List every claim the agent made (for example 'fixed the login bug', 'all tests pass', 'added tests', 'no breaking changes'). 2) For each claim, say Supported, Not supported or Can't tell, and point to the exact file, line or log that shows it. 3) Flag red flags: tests deleted, skipped or weakened, checks turned off, error handling removed, files changed that have nothing to do with the task, secrets or config edited. 4) Write a short receipt I can paste into the pull request: what was really done, what wasn't, and what to verify before merging. 5) Suggest 3 instructions I can add to my agent's rules so it stops making unverified claims (for example 'never delete or skip a test to make the suite pass; ask instead').
How to use it
- Copy the prompt above.
- Replace every [bracket] with your own details (or fill in the blanks on AgentPlaybook Real Estate, which remembers your answers for the next prompt).
- Paste it into proofslip (GitHub Action), GitHub, Claude Code and review the answer before you use it.
Works with: proofslip (GitHub Action), GitHub, Claude Code, Codex
Real-world example: x.com/kickscosmo/status/2114111189009764378
Featured in: AI prompts to summarize meetings, notes, documents and reports · AI prompts for writing SOPs, reports, policies and checklists
Before you use it: what to check
Check yourself
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
Using an AI agent? Download this prompt as a SKILL.md agent skill with these checks built in, or get the .zip for Claude.ai. How to install skills.
Want to make this prompt even better? Grade it with the free prompt grader or build your own with the AI prompt generator.
32 copy-paste agent prompts (PDF) plus 5 new prompts every week.
Get it free