DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel Phase 3b.1 — cross-chapter consistency review focused on character voice, plot, world rules, and repeated phrasing. Builds a rolling chapter summary then windows over the manuscript in 4-chapter groups, emitting per-window critique and merging into critique/consistency.md. Hands targeted fixes back to…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel Phase 2 — sequential chapter drafting from the act+scene outline. Each chapter is drafted, gated by mechanical slop detectors, manuscript-repetition scan, and a sentence-grading pass; flagged spans are surgically rewritten in place (see references/surgical-rewrite.md). State.json tracks chaptersdrafted /…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel Phase 4 — typeset and export the completed novel as PDF (via LaTeX), ePub (stdlib), and audiobook transcript chunks ready for an optional external TTS pipeline. Reads chapters/ch.md plus outline.md for the table of contents. Writes export/novel.pdf, export/novel.epub, export/audiobook/. Final phase — sets…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel Phase 1 — build world.md, characters.md, outline.md, canon.md, voice.md from a seed concept. Iterates until foundationscore (from lib/slopdetect.py) meets the adaptive threshold. Reads $AUTONOVELWORKSPACE/$NOVELSLUG/seed.txt; writes the five foundation files; updates state.json. Hands off to…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Use when creating, extending, or validating AutoNovel genre packs. Do not use for drafting, evaluating, or generating novels.
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel manual grading — user-triggered quality report. Runs every mechanical detector (slopdetect, slopngrams) plus the sentence-grading pass over a scope ('chapter 7', 'chapters 3-9', 'manuscript') and writes critique/slopreport .md: a self-contained report the user can fix by hand or paste into a stronger model.…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel Phase 3b.2 — user-triggered targeted prose review. Takes a scope (e.g. 'ch 7', 'ch 7-10', 'act 2', 'the heist arc') and emits line-level critique for that scope only. NOT auto-invoked by the pipeline. Designed for the user to sit with the agent and iterate on prose quality chapter-by-chapter or arc-by-arc.…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Autonovel Phase 3a — iterative revision via adversarial edit + reader-panel consensus + mechanical cuts. Per cycle: generate structural critique, gather 3-4 reader-persona reactions, identify mechanical fixes via lib/slopdetect.py, consolidate into a revision brief, then execute targeted rewrites. Loops 3-6 cycles…
DukeTwoCan/autonovel-agent-skills
Skill Claude CodeCodex
Entry point for the autonovel pipeline — multi-phase novel-writing skill pack. Use when the user wants to write a novel from a seed concept, resume an in-progress novel, or learn what the pipeline does. Dispatches to phase-specific skills (foundation, drafting, revision, consistency-pass, prose-review, export) based…