What problem does it solve?
Lessons learned during AI-assisted tasks are lost between sessions, forcing agents to rediscover the same fixes and repeat the same debugging loops. This Skill enforces a distillation pass after every task so corrections, workarounds, and novel solutions become reusable procedures in a persistent skill library.
Core Features & Use Cases
- Trigger-Based Extraction: Automatically runs after user corrections, debugging sessions over three iterations, undocumented workarounds, or heartbeat retrospectives.
- Classification and Routing: Decides whether new knowledge updates an existing skill, creates a new skill, goes to meta-skills, or belongs in memory as a one-off fact.
- Library Health Checks: Detects bloat (files over 500 lines), duplication across skills, and stale version references, with sidecar offload rules for scripts, examples, and lessons.
- Draft Skill Lifecycle: Flags skills created mid-task as high-priority drafts, enforces short check-in cycles, and defines promotion criteria to stable status.
- Use Case: After a long Isaac Sim debugging session where a rogue RigidBodyAPI caused physics explosions, run the distillation pass to convert that one-off fix into a general asset stability check procedure stored in the library.
Quick Start
Run a distillation pass on the task we just finished and update the skill library with any reusable lessons.
