Gen Characters
Generate or revise the character registry (characters.md) for a novel. Use during the foundation phase after world.md exists. Each character gets wound/want/need/lie, three-slider profile, distinct speech, secrets, and relationships.
Voice Discovery
Discover the voice for a specific novel during the foundation phase. Writes 3-5 trial passages in different registers, evaluates which best serves the world and tone, then fills in voice.md Part 2 with the chosen register, exemplars, and anti-exemplars.
Gen Brief
Auto-generate a chapter revision brief from structured feedback (eval JSON + reader panel JSON + adversarial cuts JSON + Opus review items). Output is a markdown brief suitable as input to gen-revision. Picks the right brief shape (compression / expansion / dramatisation / character-deepening / consistency).
Gen Revision
Rewrite a single chapter from a revision brief. Preserves voice, world, and characters from the existing draft while making the structural changes the brief specifies. Used after gen-brief and during revision cycles.
Seed
Generate fantasy novel seed concepts. Use when the user asks for novel ideas, premises, story seeds, or wants to brainstorm a new novel. Also use to riff variations on an existing seed idea.
Opus Review
Send the full manuscript to two parallel personas — literary critic + professor of fiction — for deep prose-level review. Parses the result with scripts/parse_review.py to extract numbered actionable items with severity and qualification status. Decides whether to keep revising or stop based on stopping conditions. Use after automated revision cycles plateau.
Compare Chapters
Head-to-head literary comparison of two chapters from the same novel. Picks a winner (no ties allowed), quotes the deciding passage. Used to build Elo rankings via Swiss-style tournament during revision.
Evaluate Foundation
Score a novel's planning documents (world.md, characters.md, outline.md, voice.md, canon.md) as an integrated foundation. Returns dimension scores 0-10, identifies the weakest dimension and top three improvements. Used to gate exit from the foundation phase (foundation_score > 7.5 AND lore_score > 7.0).
Evaluate Chapter
Score a single drafted chapter against the planning docs and the universal slop guardrails. Returns voice/beat/character/prose/canon dimension scores 0-10 and a top-3 revision list. Used during drafting (keep if score > 6.0, retry otherwise) and to find weakest chapters during revision.
Autonovel
Top-level orchestrator for the autonomous novel pipeline. Drives Phase 1 (foundation), Phase 2 (drafting), Phase 3a (automated revision), and Phase 3b (Opus review loop) by delegating to the seed/gen-world/gen-characters/gen-outline/gen-canon/voice-discovery/draft-chapter/evaluate-*/adversarial-edit/apply-cuts/reader-panel/gen-brief/gen-revision/opus-review skills. State persists in <novel-dir>/s…
Adversarial Edit
Adversarial editing pass — given a chapter, identify 10-20 specific cuts (FAT / REDUNDANT / OVER-EXPLAIN / GENERIC / TELL / STRUCTURAL) with exact quotes. The cut list IS the revision plan. Used in revision cycles before reader-panel.
Gen Canon
Extract every hard fact from a novel's planning documents into canon.md — the consistency database that future chapters and revisions check against. Use after world.md and characters.md exist; rerun after major changes.
Apply Cuts
Apply adversarial-edit cuts to chapter files (mechanical quote-matching removal). Wraps scripts/apply_cuts.py. Filters by cut type (OVER-EXPLAIN, REDUNDANT typically yield ~55-60% of cuts). Use after adversarial-edit during revision cycles.
Gen World
Generate or revise a novel's world bible (world.md). Use during the foundation phase, after a seed exists and voice has been at least sketched, to build out lore, magic system, geography, factions, and culture. Also use to expand a specific weak section flagged by evaluate-foundation.
Draft Chapter
Write a single full chapter of the novel from the outline entry. Use during the drafting phase. Locks to the POV, hits every outline beat, follows voice.md, and avoids the AI-fiction patterns flagged in framework/ANTI-PATTERNS.md.
Voice Fingerprint
Run the mechanical voice-fingerprint scanner across all chapters to identify outliers in sentence rhythm, vocabulary domain balance, dialogue ratio, abstract noun density, and similar metrics. No LLM call — just runs the Python script. Use during revision to find chapters that drift from the novel's voice.
Gen Outline
Build a chapter-by-chapter outline (outline.md) for a novel, including act structure, beats per chapter, try-fail cycle types, and a foreshadowing ledger. Use during foundation phase after world.md and characters.md exist.
Evaluate Full
Holistic novel-level evaluation — arc completion, pacing curve, theme coherence, foreshadowing resolution, world/voice consistency, overall engagement. Returns the weakest chapter and the top single suggestion. Use after first draft and between revision cycles.
Reader Panel
Run a 4-persona reader panel (editor, genre reader, writer, first reader) on the full novel. Spawns four subagents in parallel, each evaluating independently, then merges results into consensus + disagreement findings. Use during revision cycles to surface novel-level issues that per-chapter eval misses.