What problem does it solve? Improvements made inside a project — new skills, personas, invariants, or instructions — often stay trapped in that repo instead of benefiting the shared ai-badger framework. This Skill detects those additions, generalizes them, and opens a draft PR against the framework for human review. ## Core Features & Use Cases - Candidate Detection: detect_additions.py diffs .ai-badger/ against manifest.json and emits new and changed files as structured JSON candidates. - Generalization Workflow: Guides classification of each candidate as agnostic, generalizable, or project-specific, stripping repo names, domain terms, and absolute paths before placement. - Guarded Draft PRs: open_pr.py stages only explicitly declared paths, scans them for credential-shaped literals, and opens a draft PR via gh — never auto-merging. - Use Case: After refining a code-review persona in a scaffolded project, run this Skill to detect the change, generalize it, place it under the right stack/feature path, and open a draft PR to Arasz/ai-badger. ## Quick Start Ask the agent to feed the reusable improvements in this repo back to the ai-badger framework as a draft PR.