challenge-change
Skill958 starsChanged 14 days agoApache-2.0View on GitHub
Challenge an existing change by investigating plausible blind spots despite passing checks. Use for adversarial review, /review-pr, or an explicit challenge request through /verify-change; keep implementation and tests unchanged.
In this repository8 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- apicurio-test-qualitySkill
- implement-changeSkill
- plan-changeSkill
- sdk-updateSkill
- security-reviewSkill
- storage-variant-guideSkill
- verify-changeSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Read AGENT-VERIFICATION.md at the repository root.
- Distinctive: line 18
Use relevant specialist skills in report-only mode.
- Distinctive: line 20
Use isolated experiments and disposable data.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Challenge a Change
Read AGENT-VERIFICATION.md at the repository root. Follow its shared workflow
contract and section 4. This is a report-only pass: leave implementation and tests
unchanged, including during specialist reviews.
- Establish the complete requested diff and revision, criteria, and review level. Label inferred intent unconfirmed. Reuse existing evidence only if it matches the reviewed revision and environment; an implementation summary is not proof.
- Inspect the diff and actual assertions. Identify a plausible consequential failure that the current checks would miss. Look for weakened/deleted tests, shared implementation/test assumptions, unintended contracts/defaults, and failure paths.
- Use relevant specialist skills in report-only mode. Investigate the strongest concern with a targeted existing check or reproducible experiment where feasible. Use isolated experiments and disposable data; do not edit project code or tests.
- Distinguish observed defects, hypotheses, and unverified behavior. Prioritize actionable findings over speculative lists or style suggestions. Avoid repeating a full verification suite unless new evidence or unresolved concerns justify it.
Return the shared evidence summary, findings with file:line or method references, and exact decisions requiring human inspection. If no defect is found, state the limits of the investigation. Stop after reporting; do not fix findings or submit a GitHub review unless separately requested. Agent agreement is not human approval.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

