analytics-feedbacklisted

# analytics-feedback Turns channel analytics into pipeline behavior. YouTube Studio → Analytics → **Content** → Export → unzips to `~/Downloads/Content <range> <channel>/` with `Table data.csv` (per-video views, watch hours, CTR). This skill reads that export and re-derives which topics to chase and which to bury. ## Usage ```bash analytics-feedback.sh # newest export under ~/Downloads/Content*C0BALT_CUT*/ analytics-feedback.sh "/path/Table data.csv" # explicit file analytics-feedback.sh --force # re-learn even if the export's mtime is unchanged ``` Idempotent: records the consumed export's mtime in `work/_autopilot/analytics.csv.mtime` and no-ops on an unchanged export (so it's safe to call every autopilot tick). Drop a fresh export into `~/Downloads` and the next tick relearns. ## What it writes 1. **`.claude/skills/schedule-drip/topics.scorelist`** — preserves everything ABOVE the `# ==== AUTO (analytics-feedback) ...` sentinel (your hand-curated rules are never touched) and regenerates the block below it with data-derived `GO <pattern>` / `HOLD <pattern>` lines. Evidence (`n=`, `med=`, `ctr=`) is on `#` comment lines — never inline on a rule line, because `schedule.py` treats everything after the verdict as the regex. 2. **`.claude/skills/scout-sources/niches.txt`** — preserves your manual seed queries and appends a regenerated AUTO block of search queries built from the GO winners (e.g. a winning `black[ -]?hole` → `black hole