self-audit
Skill0 starsChanged 49 days agoMITView on GitHub
Pre-submission audit that fires right before you hand a report, analysis, or conclusion to a human or another system. Machine-checks for unbacked claims, silent "it's zero / it's stopped" misdiagnoses, and unverified API assumptions. Also triggers on "audit this" or "check for guesses".
6 other projects use a version of this.
Adopted by
Most recently taken up by
- helmut-hoffer-von-ankershoffen/heinzeltheir own version, first seen Nov 8, 2026
- 0xUrsanomics/utopia-ostheir own version, first seen Nov 1, 2026
- poa-box/poa-clitheir own version, first seen Jul 31, 2026
- faizkhairi/claude-code-blueprinttheir own version, first seen Jul 3, 2026
- Xxt-XN/agent-self-audittheir own version, first seen Jun 15, 2026
- jamestexas/agentstheir own version, first seen May 16, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- faizkhairi/claude-code-blueprint★ 73same name
- 0xUrsanomics/utopia-os★ 2same name
- helmut-hoffer-von-ankershoffen/heinzel★ 2same name
- jamestexas/agents★ 2same name
- poa-box/poa-cli★ 2same name
- Xxt-XN/agent-self-audit★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository6 other files it gives its agents.
- analysis-sweepSkill
- cloud-run-job-deploySkill
- grillingSkill
- pre-deploySkill
- systematic-debuggingSkill
- tddSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
3
Distinctive
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 7
Run this mechanically right before you submit any report, analysis, or conclusion.
- Distinctive: line 18
Never assert "does not exist / no prior art / nobody has done this".
- Distinctive: line 36
All checks pass -> no footnote needed.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Pre-Submission Self-Audit (Guess Check)
Run this mechanically right before you submit any report, analysis, or conclusion. If even one item is NG, do not submit — fix it first.
Checklist
1. Claim-to-source table
- For every number and factual claim in the output, can you name the exact source (query result / API response / file / log) where you verified it, one-to-one?
- For anything you cannot: (a) go verify it at the source, (b) cut it, or (c) tag it
[unverified].
2. "It's zero / it's missing / it's stopped" assertions (the most frequent failure mode)
- "The data is zero" has two readings: actually zero and the fetch failed and it looks like zero. Did you check both?
- "It's stopped" has two readings: actually halted and the copy I'm looking at is stale. (Real example: a "backup stopped" alarm that was just a local checkout that had not pulled.)
- Never assert "does not exist / no prior art / nobody has done this". Stop at "not found within the range I searched" — proving nonexistence is impossible.
3. External API / library behavior
- Any place you assert a field name, function signature, or behavior: is it confirmed against official docs? Tag unconfirmed spots
[unverified].
4. Explicit split between fact and guess
- Is every claim classified as either "verified (source named)" or "guess"? Do not leave mushy hedges ("should be", "probably", "seems to") standing in for one or the other.
5. Per-recipient rules
- Different audiences need different framing. Keep a short table of your own recipient rules here and check against it before sending. Template:
<recipient A>: <the specific basis / unit / framing they always expect; the mistake they always catch><recipient B>: <plain text only, no Markdown tables, no nesting — paste-safe><daily report>:<the fixed format your reader expects>
6. Three disconfirming questions
- List three ways "this conclusion could be wrong" and show you closed each one. Anything you cannot close, state as a stated limitation.
Output
- All checks pass -> no footnote needed. Submit as-is.
- Something was NG -> fix before submitting, and add one line noting what you changed (e.g. "note: added source verification for X during pre-submission audit").
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.