feature-minerlisted

# Feature Miner Mine coding agent session history (via `cass`) to discover how AI agents interact with documentation in real projects. Extract patterns, pain points, and feature ideas for improving autodoc. ## Purpose Autodoc manages documentation for AI coding agents. Real-world session data reveals how agents actually use (or fail to use) docs — surfacing gaps, friction, and opportunities that code review alone cannot. ## Prerequisites - `cass` CLI installed and indexed (`cass health` returns healthy) - If index is stale, run `cass index` first ## State Tracking Track the last run timestamp in `.autodoc/feature-miner-state.json`: ```json { "last_run": "2026-03-12T00:00:00", "last_session_ts": 1773323229563 } ``` On first run, initialize this file. On subsequent runs, use `--since` with the `last_run` date to avoid re-scanning old sessions. After completing a mining run, update `last_run` to the current date and `last_session_ts` to the highest `created_at` value seen. Read this state file at the start of every run. If the file exists and `last_run` is recent (within 1 day), inform the user and ask whether to re-scan or only look at new sessions. ## Mining Workflow ### Phase 1: Search for Patterns Run the search script to gather raw results across multiple query categories: ```bash bash ${SKILL_DIR}/scripts/mine-sessions.sh [--since YYYY-MM-DD] ``` The script runs parallel cass searches across these categories: 1. **Doc discovery** — agents trying to find,