| name | critique |
| description | Critique a document for logical fallacies and weaknesses. Identifies issues and provides actionable fixes. |
| disable-model-invocation | true |
Critique
Analyze a document for logical fallacies and structural weaknesses. Every issue
must include a recommended fix.
Process
-
Get the file. If no path was given, ask:
Which file do you want me to critique?
-
Read the file in full.
-
Scan for logical fallacies. Not exhaustive — flag any fallacy you spot,
listed or not:
- Straw man — misrepresenting a position to attack it
- False dichotomy — only two options when more exist
- Appeal to authority — true because an authority said so, no evidence
- Slippery slope — unlikely chain of consequences, unjustified
- Circular reasoning — conclusion used as a premise
- Ad hominem — attacking the person, not the argument
- Red herring — irrelevant points to distract
- Hasty generalization — broad conclusion from limited evidence
- False cause — correlation treated as causation
- Moving the goalposts — changing criteria after the fact
- Equivocation — a term with shifting meaning
- Appeal to emotion — emotion substituted for evidence
- Bandwagon — true because many believe it
- Begging the question — conclusion assumed in the premise
- Tu quoque — deflecting criticism onto the accuser's behavior
-
Critique the document. Evaluate:
- Argument structure — claims supported? gaps in reasoning? conclusions
follow from premises?
- Evidence quality — sources cited? evidence relevant and sufficient?
statistics used correctly?
- Assumptions — what unstated assumptions exist? reasonable?
- Completeness — counterarguments addressed? important perspectives
missing?
- Clarity — terms defined? ambiguous or vague where precision matters?
- Consistency — self-contradiction? later sections conflict with earlier
claims?
-
If the document is an implementation plan (e.g. under docs/plans/, or
structured with tasks, File Changes, and Verification sections),
additionally evaluate:
- Scope integrity — is every task within the stated goal, or do some
drift into adjacent refactors, renames, or cleanup the objective doesn't
require?
- File enumeration — does the File Changes table list every file the
tasks imply touching? Flag tasks whose implementation needs unlisted files.
- Task testability — is each task independently verifiable? Flag "improve
X" / "clean up Y" / "refactor Z" tasks with no pass/fail signal.
- Verification concreteness — does the verification phase name actual
commands, test paths, or observable outcomes? Flag vague "run the tests" or
"confirm it works".
- Unstated dependencies — tasks implicitly depending on another (shared
file, new type, upstream contract) without declaring it?
- Edge cases — unaddressed failure modes: error paths, partial failures,
concurrent callers, rollback, data migration of existing state.
These checks apply only to plans. Do not manufacture plan issues on
non-plans (blog posts, research docs, design notes).
Output
Two sections plus a summary. For every issue, quote the passage and give a
concrete fix.
Logical Fallacies
Per fallacy (if none, say so explicitly):
**{Fallacy Name}**
> {Quoted passage}
Problem: {Why it's a fallacy — one or two sentences.}
Fix: {Specific rewrite or approach to eliminate it.}
Document Weaknesses
Per weakness:
**{Category}** — {Brief title}
> {Quoted passage}
Problem: {What is weak and why it matters.}
Fix: {Concrete recommendation — rewrite, add evidence, restructure, etc.}
Summary
Close with: how many fallacies were found, the most significant weaknesses, and
the single highest-impact improvement the author could make.
Guidelines
- Be thorough but fair. Flag real issues, not stylistic preferences.
- Quote the source so the author can locate each issue.
- Fixes must be actionable — a specific rewrite or concrete next step, not "make
this better".
- If the document is well-constructed, say so. Don't manufacture issues.
- Don't rewrite the whole document. Focus on the weakest points.