critical-qa
Skill2 starsChanged 5 months agoMITView on GitHub
Self-adversarial Q&A on a code artifact. Run hostile-reviewer questions across ≥3 distinct categories (mechanism / edge / confidence / failure / counterfactual / adversarial / composition / honesty), attempt honest answers, surface weak spots. Per [P·self-adversarial-qa] (a partner chat) origin). Use BEFORE declaring code complete; sibling of /classify but on the quality axis not the cost axis.
In this repository7 other files it gives its agents.
- classifySkill
- house-styleSkill
- literature-loopSkill
- marginal-contributionSkill
- md-to-pdfSkill
- ship-webSkill
- trp2Skill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 51
no softball questions (anti-pattern: "is this code clean?
- Distinctive: line 52
if answer is "I don't know" — surface as research-needed
- Distinctive: line 53
if N consecutive rounds find nothing new → ship per [P·discovery-ceiling]
- Distinctive: line 59
Use this on YOUR OWN code.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
/critical-qa — self-adversarial Q&A
You are the implementer of a code artifact. A hostile reviewer is about to arrive. Pre-empt them by asking the questions YOU would ask if you were the reviewer.
Input: $ARG = code area / file / function / hook to interrogate.
Question categories (pick ≥3 distinct)
- Mechanism transparency — Who/what is literally doing X? Name the answer.
- Edge cases — Empty / massive / adversarial / concurrent input?
- Confidence exposure — Does output convey its own confidence? Silent failures?
- Failure modes — What fails? Loud or quiet?
- Counterfactual — If my central assumption is wrong, would I notice?
- Adversarial — How would someone break this?
- Composition — Integration surface with existing system?
- Honesty — Verified or extrapolated?
Output shape
## /critical-qa on <$ARG>
### R1 [category]
Q: <hostile question>
A: <honest answer; if uncertain, say so>
→ <weak spot found OR confirmed-ok>
### R2 [different category]
...
### R3 [different category]
...
## Findings
- <fixable now>
- <annotate as known weakness>
- <surface to operator for decision>
## Verdict
[ship | revise | research-needed | escalate]Rules
- ≥3 rounds across ≥3 distinct categories
- no softball questions (anti-pattern: "is this code clean? yes")
- if answer is "I don't know" — surface as research-needed
- if N consecutive rounds find nothing new → ship per [P·discovery-ceiling]
- compose with /classify (cost-axis) for spawn decisions
Sibling
- Code-side companion to [P·adoption-roleplay] which is partner-artifact-side.
- Use this on YOUR OWN code; use AdoptionRoleplay on artifacts going to partners.
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.

