kaozheng
Skill8 starsChanged 7 months agoApache-2.0View on GitHub
Use when a historical paper or final draft needs evidential scrutiny: warrants, citations, source rank, truncation risk, or argument structure must be audited rather than stylistically interpreted.
In this repository16 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .cursor/rules/collate.mdcCursor rule
- chunqiuSkill
- constellatioSkill
- diff-reviewSkill
- mp-formatSkill
- ocr-runSkill
- paper-summarySkill
- prep-scanSkill
- prometheusSkill
- proofreadSkill
- real-thesisSkill
- setupSkill
- to-docxSkill
- visual-previewSkill
- xray-paperSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 37
If only raw.md exists, note that the text has not yet been proofread and mark the report accordingly.
- Distinctive: line 42
Read the argument in layers:
- Distinctive: line 71
Never modify the source text.
- Distinctive: line 72
Do not confuse disagreement with disproof.
- Distinctive: line 74
If verification is impossible from available material, mark it as unverified rather than pretending certainty.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
kaozheng
The work this skill is for
This is not typo-catching. That is proofread.
kaozheng reads in the Qian-Jia evidential tradition: not asking whether the
paper sounds plausible, but whether its evidentiary bridge bears weight.
The source may be real and the bridge still false. A citation may be accurate
and still be pressed into service beyond what it can support.
What it audits
For a post-OCR, post-proofread paper, test:
- whether the claim is actually supported by the data cited
- whether quotations are complete enough to preserve original sense
- whether cited authorities are first-rank, second-rank, or hearsay
- whether the warrant is visible and defensible
- whether the paper leans on a solitary witness (
孤证不立)
Input
Target: $ARGUMENTS
- workspace directory → read
<ws>/final.md .mdfile → read it directly- empty → ask what to read
If only raw.md exists, note that the text has not yet been proofread and mark
the report accordingly.
Method
Read the argument in layers:
- Extract the paper's main claims.
- Pair each claim with its offered evidence.
- Identify the warrant that connects the two.
- Audit quotations and citation framing:
- what is quoted
- what is omitted
- what rank the source holds
- whether a footnote quietly carries more force than the body text
- Flag cardinal errors first; breadth comes second.
The framework is a scale, not a blade. Do not deform the paper merely to force it into Toulmin language.
Output
- workspace mode →
<workspace>/analysis/{stem}_kaozheng.md - single-file mode →
analysis/{stem}_kaozheng.md
Report structure:
- Argument skeleton — claim / data / warrant / backing / qualifier / rebuttal
- Citation audit table — original fragment → suspected source → verifiable? → truncated?
- Cardinal errors — one to three, ranked by severity, with repair suggestions
- Suspicious but unverified — handed off for further checking
Guardrails
- Never modify the source text.
- Do not confuse disagreement with disproof.
- One decisive structural error outweighs twenty decorative observations.
- If verification is impossible from available material, mark it as unverified rather than pretending certainty.
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.

