Multi-agent AI novel writing system with Sequential Competition, Mix Edition, and council-based quality gates.
Omni Forge Novel is an orchestration skill for AI-assisted novel writing. It runs multiple AI agents against the same scene or chapter beat, evaluates outputs through a council scoring system, and selects (or mixes) the best result — producing consistently higher-quality prose than single-pass generation.
Multiple AI agents each write the same passage independently. Outputs are scored across dimensions (prose quality, pacing, consistency, originality) and the winner advances to the next stage.
Instead of picking a single winner, the system extracts the strongest elements from each candidate and merges them into a hybrid output — combining narrative momentum from one draft with character voice from another.
A dedicated evaluation agent (or panel) scores each candidate on a 0–10 rubric. Only outputs scoring ≥ 7 pass the gate and move forward.
- Writing chapters or scenes with high creative stakes (climax, twist, opening, ending)
- Breaking out of repetitive AI prose patterns
- Generating multiple stylistic variants before committing to a direction
- Long-form novel projects requiring consistent quality across 50–100+ chapters
| File | Purpose |
|---|---|
SKILL.md |
OpenClaw skill entrypoint |
forge-workflow.md |
Full pipeline workflow documentation |
checklist.md |
Pre/post-forge quality checklist |
diagnostics.md |
Debugging stuck agents, stalled outputs |
Beat prompt → N agents write independently
↓
Council scores each output (0–10)
↓
Score ≥ 7? ──Yes──► Select best / Mix edition
──No───► Retry with adjusted prompt
↓
Final output → Novel Guardian scan → FINAL
- novel-guardian — Continuity + consistency checker
- novel-master — 5-layer chapter QA system
- novelcore-ai — Structured beat prompting (4-block, 3-variant)
- forge-novel-guard — Pre/post-forge verification scripts
Part of the NachaFromMars OpenClaw skill ecosystem.