flywheel-mine-patterns

What problem does it solve? Repeated prompts, tool sequences, and error-fix loops accumulate invisibly across agent coding sessions, and teams never codify that tacit knowledge into reusable skills or pipeline improvements. ## Core Features & Use Cases - Ritual detection: Groups user prompts repeated 5+ times across sessions to surface workflows worth automating. - Tool sequence mining: Extracts recurring tool n-grams (e.g., Read → Edit → Bash) appearing across 5+ distinct sessions. - Error-fix loop discovery: Finds failure markers followed by successful retries and pairs them as codification candidates. - Classification and bead creation: Buckets each candidate as a new skill scaffold, an existing-skill refinement, or a pipeline tweak, then creates br beads only after user confirmation. - Use Case: After a long sprint, run the miner over the last 7 days of cass-indexed sessions to get a numbered, evidence-backed list of automation candidates and land the chosen ones as tracked beads. ## Quick Start Ask the agent to mine the last 7 days of cass session history in the current workspace and propose recurring workflows as new skill candidates.