curate-delta

Playbook curation synthesizes failure reflections into structured delta proposals—new guidance bullets, corrections, edits, and deprecations—that generalize learning beyond the immediate task. This Skill applies ACE Curator workflows to Reflector output and execution feedback to generate deterministic JSON deltas that prevent recurring failures.

Filter vague or redundant guidance by validating evidence and tagging metadata on curated deltas. This ensures only high-confidence, actionable patterns propagate into your playbook, maintaining consistency across automation rules.

Yes. When your AI assistant encounters new error patterns, delta curation structures the learning into precise, actionable guidance with evidence and reproducible metadata, eliminating manual analysis of task failures.

Delta output includes new_bullets (fresh guidance), counters (exception handling), edits (refinements to existing rules), merges (consolidated patterns), and deprecations (obsolete guidance), all in validated JSON with taggable metadata for reproducibility.

Inputs require Reflector insights, execution feedback, and current playbook state. The Skill generates deterministic JSON deltas, so structured input from Reflector workflows ensures precise, actionable curation output.

Automated delta synthesis from task reflections feeds quality-filtered, evidence-backed guidance directly into playbooks, accelerating AI model improvement without manual review overhead in context-engineering pipelines.