autonovel
CLAUDE.md1 starsChanged 7 months agoMITView on GitHub
This file is auto-loaded by Claude Code when the repo is opened. It documents the conventions a contributor (or Claude) needs to know before touching the pipeline. An autonomous fantasy-novel pipeline implemented as Claude Code skills, subagents, and slash commands. The user runs /autonovel <tag> and the orchestrator drives foundation → drafting → revision → Opus review. All run state lives in novels/<tag>/state.json. Resume via /autonovel-resume <tag>. There is no global state outside a tag directory. Phase transitions write snapshots…
- Reads credentials
- Deletes or force-pushes
The file
We split this CLAUDE.md into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 28
All run state lives in novels/<tag>/state.json.
- Very specific: line 55
Read state via scripts/state.py.
- Distinctive: line 56
Use snapshots before any rewrite that touches more than one chapter.
- Very specific: line 57
Keep framework/*.md immutable inside a single run.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
autonovel — Claude Code project notes
This file is auto-loaded by Claude Code when the repo is opened. It documents the conventions a contributor (or Claude) needs to know before touching the pipeline.
What this repo is
An autonomous fantasy-novel pipeline implemented as Claude Code skills,
subagents, and slash commands. The user runs /autonovel <tag> and the
orchestrator drives foundation → drafting → revision → Opus review.
Directory contract
| Path | Role | Edit policy |
.claude/skills/ | LLM-driven pipeline steps | Edit when changing pipeline behavior |
.claude/agents/ | Persona subagents (panel, critics) | Edit when tuning persona voice |
.claude/commands/ | Slash commands (/autonovel) | Edit when changing user-facing entry points |
framework/ | Immutable craft references | Do not edit mid-run — skills load these as ground truth |
templates/ | Per-novel skeletons | Edit to change what a fresh run starts with |
scripts/ | Mechanical Python helpers (no LLM) | Pure functions; no API keys, no network |
novels/<tag>/ | One per novel run | Gitignored; state lives here |
examples/ | Worked examples (committed) | Demo runs only — do not write new runs here |
State
All run state lives in novels/<tag>/state.json. Resume via
/autonovel-resume <tag>. There is no global state outside a tag directory.
Snapshots, not git
Phase transitions write snapshots to novels/<tag>/.snapshots/. "Discard"
means restore from snapshot — never git reset --hard. The repo's git
history tracks the pipeline, not any individual novel run.
Models
Skills assume an Opus-class model is available for Phase 3b review and a Sonnet-class model for drafting/evaluation. On a Haiku-only plan expect degraded prose quality and weaker review signal. The harness (Claude Code) chooses the model — there is no model-name branching in skills.
API keys
Claude Code provides the LLM, so ANTHROPIC_API_KEY is not needed for
the pipeline itself. .env.example is preserved only for users who want to
run the standalone Python scripts in scripts/ against the Anthropic API
directly (none currently require it; reserved for future helpers).
Conventions when editing skills
- Skills must be idempotent on re-run — the orchestrator may invoke them twice on resume.
- Read state via
scripts/state.py; never parsestate.jsonad-hoc. - Use snapshots before any rewrite that touches more than one chapter.
- Keep
framework/*.mdimmutable inside a single run; bump them between runs if craft references evolve.
What's intentionally out of scope
Art generation, audiobook synthesis, ePub export, LaTeX typesetting,
landing-page generation. The legacy Python tooling for those lived in a
separate inbox/autonovel/ directory in the original repo and is not
ported here.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.