Do an editorial review of the given post or article. Report findings. Don't edit
the text.
Target
jj show --git
Arguments: $ARGUMENTS
Review the target named in the arguments if given. Otherwise review the prose
changed in the current commit shown above. If there are no arguments and the
commit has no changes, ask the user what to review and stop.
Workflow
- Read the whole piece once as a reader would, before critiquing
- Determine the intended audience and the point the piece argues for. The
passes judge the piece against these
- Apply the passes in "Passes"
- If the passes find nothing significant, tell the user the piece is in good
shape and stop. Don't manufacture findings
- Report findings in order of importance, where a structural problem outranks a
line-level one. For each: what the problem is, where it occurs, why it
weakens the piece, and a concrete suggestion
- Suggest
/humanize if the main weakness is the prose itself rather than
substance or structure
Passes
- Thesis: can you state the piece's point in one sentence? Does every section
serve it? Flag sections that don't
- Audience fit: assumed knowledge, jargon, and depth match the intended reader.
Flag explanations the audience doesn't need and gaps they do. If the intended
reader can't be inferred, report that
- Structure: does the order of sections build the argument? Would a reader
skimming the headings follow the argument?
- Evidence: are claims supported by examples, data, or reasoning? Flag bare
assertions a skeptical reader would push back on
- Counterarguments: are obvious objections acknowledged? Flag the strongest
unaddressed one
- Opening: does the first paragraph earn the read, or does it warm up? Flag
burying the lede
- Conclusion: does it land the thesis, or trail off into summary and filler?
- Title: accurate, specific, and matching what the piece delivers, not
clickbait or vague
- Redundancy: points made more than once without adding anything. Recommend
which occurrence to keep
- Tone: consistent register throughout. Flag lurches between formal and
casual, or confident and hedging
These aren't exhaustive. Flag a problem no pass covers when the same reasoning
applies.
Guidelines
- Critique the piece the author is writing, not the one you would write
- Be specific: quote or point to the offending passage, never "the middle
section feels weak"
- Don't flag voice or stylistic choices that are deliberate and consistent
1---2name: editorialize3description: Do an editorial review of a post or article and report findings ordered by importance.4---56Do an editorial review of the given post or article. Report findings. Don't edit7the text.89# Target1011```!12jj show --git13```1415Arguments: $ARGUMENTS1617Review the target named in the arguments if given. Otherwise review the prose18changed in the current commit shown above. If there are no arguments and the19commit has no changes, ask the user what to review and stop.2021# Workflow22231. Read the whole piece once as a reader would, before critiquing242. Determine the intended audience and the point the piece argues for. The25 passes judge the piece against these263. Apply the passes in "Passes"274. If the passes find nothing significant, tell the user the piece is in good28 shape and stop. Don't manufacture findings295. Report findings in order of importance, where a structural problem outranks a30 line-level one. For each: what the problem is, where it occurs, why it31 weakens the piece, and a concrete suggestion326. Suggest `/humanize` if the main weakness is the prose itself rather than33 substance or structure3435# Passes36371. Thesis: can you state the piece's point in one sentence? Does every section38 serve it? Flag sections that don't392. Audience fit: assumed knowledge, jargon, and depth match the intended reader.40 Flag explanations the audience doesn't need and gaps they do. If the intended41 reader can't be inferred, report that423. Structure: does the order of sections build the argument? Would a reader43 skimming the headings follow the argument?444. Evidence: are claims supported by examples, data, or reasoning? Flag bare45 assertions a skeptical reader would push back on465. Counterarguments: are obvious objections acknowledged? Flag the strongest47 unaddressed one486. Opening: does the first paragraph earn the read, or does it warm up? Flag49 burying the lede507. Conclusion: does it land the thesis, or trail off into summary and filler?518. Title: accurate, specific, and matching what the piece delivers, not52 clickbait or vague539. Redundancy: points made more than once without adding anything. Recommend54 which occurrence to keep5510. Tone: consistent register throughout. Flag lurches between formal and56 casual, or confident and hedging5758These aren't exhaustive. Flag a problem no pass covers when the same reasoning59applies.6061# Guidelines6263- Critique the piece the author is writing, not the one you would write64- Be specific: quote or point to the offending passage, never "the middle65 section feels weak"66- Don't flag voice or stylistic choices that are deliberate and consistent