daily-forestlisted

# Daily Forest. Forest-Level Signal Synthesis ## Why this exists Signal harvesting (chat channels + RSS + email = 30-80 signals/day) accumulates thousands of Signal/ pages in the knowledge graph. Without sustained reading, the substrate becomes a searchable headline pile. useful for keyword recall, useless for spotting **patterns only visible across time**: themes converging, narratives going quiet, an authority's language shifting tone, cross-domain cross-stitches. The forest is what you miss when you only see trees. This skill is the daily forest-scan. ## When this fires - **Scheduled**: daily 21:30 (local) cron (buffer after the evening knowledge-graph sync). - **Manual**: `/daily-forest` or natural-language ("read me today's forest", "what's the synthesis from today's signals"). ## Inputs (read in this order) 1. **Today's harvest**: `$AGENT_SCRATCH_DIR/today-signals.json`. structured summary, 24-30 signals typical. 2. **Yesterday's digest**: `outputs/raw/daily-forest/YYYY-MM-DD.md` (yesterday's local date). Skip if missing. 3. **Last-7-days digests**: `ls -t outputs/raw/daily-forest/*.md | head -7`. for the silence-detection pass (themes hot in last 7d but absent today). 4. **Full source articles for the interesting signals**: for the HIGH-scored signals (up to 8), FETCH AND READ THE ACTUAL ARTICLE via its `source_url` / `urls` (WebFetch), not just the harvester's headline/summary. No url (chat-only) = fall back to the knowledge-graph Signal/ page (`$AGENT_GRAPH_DI