// Adversarial reading methodology for narrative fiction: find what doesn't work, not confirm what does. Focus-area driven with dedicated resources per area. Use when reviewing drafts, evaluating prose quality, or assessing changes at any stage.
Prose-level immersion patterns for narrative fiction. Use when writing or revising prose: the sentence-level and paragraph-level craft that pulls readers into the story. Project-specific voice comes from style files passed alongside this skill.
Arc structure, narrative design, and pacing at multiple scales: saga, arc, chapter, scene. Use when structuring story at any level, planning arcs, designing chapter outlines, or evaluating whether narrative structure serves the story's goals.
Team composition for writing workflows: which agents to spawn, how many, what focus areas to assign, and how to scale effort. Use when composing critic panels, dispatching researchers, staffing draft/revise loops, or setting up brainstorm fan-outs.
Story brainstorming capture: minimal notes that preserve creative freedom. Use when exploring narrative ideas, discussing characters, planning chapters, or thinking through story possibilities.
Logging and referencing writing issues: craft problems, tics, inconsistencies, and structural concerns found during analysis, critique, or review. Use when an agent identifies something worth tracking beyond a single critique report: repeated tics across chapters, inconsistencies that affect multiple scenes, structural problems that need the author's attention, or patterns that should be fixed in revision.
Load when producing any written artifact for humans.
| name | prose-critique |
| type | reference |
| description | Adversarial reading methodology for narrative fiction: find what doesn't work, not confirm what does. Focus-area driven with dedicated resources per area. Use when reviewing drafts, evaluating prose quality, or assessing changes at any stage. |
| model-invocable | false |
Find what doesn't work. The writer already believes their draft works: challenge that assumption. A critique that says "well done" without digging is worse than no critique, because it creates false confidence.
Your prompt specifies a focus area. Go deep on the assigned focus rather than skimming everything. Each focus area has a dedicated resource with detailed guidance:
resources/structure.md: plot logic, pacing, scene necessity, stakes, setup/payoffresources/character.md: motivation coherence, arc progression, relationship dynamicsresources/voice.md: dialogue quality, POV consistency, subtext, voice driftresources/prose.md: line-level quality, rhythm, clarity, repetition, show vs tellresources/continuity.md: facts, timeline, geography, character stateRead the relevant resource when assigned that focus. If no focus is specified, assess the draft yourself: figure out which dimensions matter most for this piece, read those resources, and focus there.
Even with an assigned focus, flag issues outside it if they're clearly serious. A voice reviewer who notices a plot hole should say so.
Good findings share these qualities:
Specific. Reference the chapter, scene, paragraph, or line. "The pacing has issues" is not a finding.
Reasoned. Explain why it matters, not just that it exists. A POV break is only interesting if you can describe what it costs: reader trust, immersion, character distinction.
Directable. The writer should know what to do after reading your finding. If the fix isn't clear, say what investigation or decision is needed.
Non-obvious. Spell-check already caught the typos. You're here for things that require understanding context, intent, and interaction between story elements.
Make it obvious which findings are serious and which are minor. The orchestrator or author triaging your findings has context you don't: they know what's intentional, what's set up for later, what's a known compromise. Give them a clear signal about severity.
Lead with the things that damage the reading experience: broken causation, character inconsistency, lost tension, confused POV. Let the smaller observations follow. Only flag issues you can tie to a concrete reader cost.
Think about how the prose fails, not how it succeeds:
Don't be adversarial for its own sake. If a section is genuinely strong, you can note it briefly, but earn that conclusion by actually reading critically, not by wanting to be nice.
Early draft: focus on structural and character issues. Line-level prose quality doesn't matter if the scene shouldn't exist or the character motivation is broken.
Mid-stage draft: structural foundations should be solid. Focus on voice, pacing, and how scenes connect.
Late draft: structure and character are committed. Focus on prose quality, line-level rhythm, word choice, and polish.
Don't spend time on prose-level polish of a scene that has structural problems. Fix the bones before the skin.
Open with a brief overall assessment: what's the big picture for this draft? Then walk through findings grouped by severity or by theme, whichever tells a clearer story. End with your verdict: what's the most important thing to address, and what's the one change that would improve this draft the most?
In multi-critic workflows (fan-out pattern), keep your report focused on your assigned area. The orchestrator synthesizes across critics: you go deep, not broad.
A bundled script measures mechanical prose properties: sentence length distribution, opener variety, dialogue ratio, repetition, pronoun distribution. These are quantitative signals useful for comparing a draft against the project's own baseline. Read-only agents (critic, continuity-checker) should request this data from the orchestrator; agents with bash access can run it directly:
uv run resources/analyze.py <file.md> [window_size]
resources/antipatterns.md: AI writing antipatterns, categorized as research-backed vs community folkloreresources/baseline.md: establishing a project baseline and comparing drafts against it