Grizzly review
Comprehensive quality review of one or more chapters. Use when the user asks for a full review, quality check, violation scan, or pre-publication pass. Runs five passes with severity-tiered findings. For "why is this boring" use grizzly-audit instead.From its SKILL.md
npx -y skills add HarishDvs/Grizzly --skill grizzly-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.2 KB, as published. Nobody here has run it
Grizzly Review — The Five-Pass Review
A structured review with severity-tiered findings, modeled on code-review culture: honest, specific, no praise padding. Scope is whatever the author names (a chapter, a range, the manuscript).
The five passes
Run in this order; report in this order. Where the harness supports parallel sub-agents, passes 1-4 can fan out (one agent per pass) and pass 5 synthesizes; otherwise run sequentially.
- Continuity (against the codex and source prose): events, facts, injuries, possessions, names, timeline. Verify apparent card conflicts in the manuscript before calling them findings; a stale card is memory debt, not a chapter error. Also check threads the chapter was supposed to advance.
- Pacing (against NOVEL.md format and any genre pack): scene lengths, where the chapter sags, payoff cadence, hook position. Serial and trad chapters obey different physics; use the declared format.
- Voice drift (against VOICE.md): passages that don't sound like the anchor lines. Quote the off-voice line, name what drifted (rhythm, register, narrator distance). This pass exists because long projects drift slowly and the author can't hear it from inside.
- Hygiene (against decks/blacklist.md + decks/register-tense.md + STANDARDS.md + VOICE.md): tells, grammatical-register and tense slips, and standards violations, each with line reference. Counts alone are useless; every instance gets a location. A register slip (trailerese maxim, present-tense leak in past narration) is severity minor unless sustained across the chapter, then major.
- Delivery (against decks/delivery-pass.md): the landing of each scene, named or found missing. The benchmark comparison where a beat underperforms.
Severity tiers
- blocker — breaks the book: continuity contradiction, a scene that defeats an established plot point, a preserve-line altered.
- major — hurts the chapter: buried landing, dead hook in a serial, sustained voice drift, a sagging middle.
- minor — hurts the line: individual tells, standards violations.
- nit — taste, flagged only if cheap to fix. Say "nit" so the author can skip.
Report format
Findings grouped by severity, then by position. Each: location, what, why it matters, and (for minor/nit) the one-line proposed fix. Blockers and majors get no inline fixes; they get a recommendation and a routing (grizzly-edit for lines, grizzly-plan for structure).
End with: the verdict in two sentences, the three highest-leverage actions, and what was checked and found clean (one line; so silence isn't ambiguous).
Rules
- Never start fixing during a review. Review, report, stop.
- If the author asks for a review of their own favorite chapter, review it exactly as hard. Calibrated honesty is the product.
- A finding you cannot locate precisely is not a finding; drop it or go verify.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most review quality skills give
Counted across 1,273 of the 2,403 authors here whose files we hold, read 2026-09-06
- Ask one question at a timein 63 of 1273, across 62 files
- Provide a recommended answer for each questionin 47 of 1273, across 45 files
- Rank findings by severityin 44 of 1273
- Use parameterized queries for database accessin 38 of 1273, across 20 files
- Validate all user input with schemasin 33 of 1273, across 15 files
- Store secrets in environment variablesin 32 of 1273, across 14 files
- Explore the codebase to answer questionsin 31 of 1273, across 29 files
- Store tokens in httpOnly cookiesin 30 of 1273, across 12 files
- Implement rate limiting on API endpointsin 30 of 1273, across 12 files
- Sanitize user-provided HTMLin 29 of 1273, across 11 files
- Return generic error messages to usersin 28 of 1273, across 10 files
- Cite file and line for every findingin 28 of 1273, across 25 files
Said here and by no other author read
- Run five review passes sequentially
- Verify continuity against codex and source
- Check pacing against format and genre
- Identify voice drift against anchor lines
- Flag hygiene issues with line references
- Evaluate scene delivery against benchmarks
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.