auctra-novel-optimization-loop

Converts Auctra novel review results and feedback into revision queues and rule proposals.

Updated Aug 3, 2026
One-click install
npx skills add https://github.com/yeisme/auctra-novel-skills --skill auctra-novel-optimization-loop-yeisme
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: auctra-novel-optimization-loop
Source: https://github.com/yeisme/auctra-novel-skills/tree/main/auctra-runtime/auctra-novel-optimization-loop
Command: npx skills add https://github.com/yeisme/auctra-novel-skills --skill auctra-novel-optimization-loop-yeisme

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Chinese novel projects managed through Auctra often stop at generating a single candidate draft, leaving review findings, user feedback, and repeated defects untracked. This Skill closes the loop by turning review results, chapter defects, context packs, and ledger deltas into an actionable revision queue, rule proposals, and a next-run optimization handoff. ## Core Features & Use Cases - Defect Attribution and Register: Classifies every issue into dimensions like context_missing, continuity_drift, style_ai_taste, or hook_pacing, each with evidence, reader impact, minimal fix, and owner skill. - Prioritized Revision Queue: Orders tasks by blocking, continuity, style, and deferred status so reader promises and chapter-end hooks are fixed before local polish. - Repeated-Issue Rule Proposals: Tracks recurring problems and proposes rule updates only after a threshold of three occurrences, always requiring user confirmation. - Auctra-Safe Handoff: Suggests real Auctra CLI commands for review actions and material storage without hand-writing structured state under .auctra/. - Use Case: After a chapter candidate is rejected for weak hooks and character state drift, run this Skill to produce a defect register, a revision queue, a next-context patch, and the exact Auctra commands to record the outcome. ## Quick Start Use the Auctra novel optimization loop to analyze this chapter's review report and user feedback, then produce a defect register, revision queue, and next-run context patch.

Frequently Asked Questions about auctra-novel-optimization-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I turn novel chapter review feedback into a revision plan?▼

Feed the review report, candidate draft, and user feedback into this Skill to get a defect register with evidence and minimal fixes, plus a revision queue ordered by blocking, continuity, and style priority. Each defect is attributed to a category like continuity_drift or hook_pacing.

How does the optimization loop handle repeated writing issues?▼

Repeated issues are counted across runs; below three occurrences they are only recorded. At three occurrences the Skill outputs a rule proposal targeting the relevant skill or reference, but the update is only suggested after explicit user confirmation.

Can this Skill accept or reject Auctra review items automatically?▼

No. It only suggests real Auctra CLI commands such as auctra review reject or auctra review partial with reasons and diffs. It never auto-accepts, auto-rejects, or hand-writes structured state under .auctra/.

What inputs does the Auctra novel optimization loop need?▼

It needs the Auctra project path, review queue items, candidate drafts, review reports, chapter acceptance results, user feedback, context packs, and ledger delta suggestions. If Auctra outputs are missing, it lists suggested commands instead of claiming to have read the queue.

What are the limitations of this optimization workflow?▼

It cannot modify skill rules without user confirmation, cannot execute Auctra review decisions itself, and must not fabricate platform or reader data. Missing CLI capabilities are reported as product gaps rather than worked around by editing state files.