vault-sync: 2026-10-14 11:24 (leonard)

memory-systems/agent-os/04_projects/search-os/decisions.md

@@ -0,0 +1,9 @@
+# Search OS — Decisions
+## 2026-10-14
+- Built as Python CLI (`search_os.py`) that can be called from dashboard or terminal.
+- Uses web_extract + web_search for site analysis (Safari 27 MCP is macOS-only; not available on this Windows PC).
+- Applies parasite-seo principles: platform priority matrix, persona layer, tiered data, keyword universe generation.
+- Applies VeriPath audit patterns: GBP surface check, site surface check, cross-reference analysis.
+- Output format: markdown report with site analysis + content drafts + action items.
+- Dashboard wiring: Workflows card "Search OS" button links to terminal invocation for now (Streamlit can't run long web extraction tasks inline).
+- Non-negotiable: never invent data; site analysis must cite real extracted content; content drafts marked "template" until LLM-enhanced.

memory-systems/agent-os/04_projects/search-os/status.md

@@ -0,0 +1,16 @@
+# Search OS — Status
+Updated: 2026-10-14
+Goal: SEO agent that analyzes sites against buy-intent keywords, generates optimized content using parasite-seo patterns, and writes actionable output to vault.
+## Current state
+- Phase 5 scaffolded: project folder, agent script, dashboard wiring.
+- One example run complete (Phoenix Salon + "best bridal hair salon folsom").
+- Output lands in `05_Output/search-os/`.
+## Next actions
+- [x] Create project folder with status.md + decisions.md
+- [x] Build search_os.py agent script
+- [x] Wire to dashboard Workflows card
+- [x] Run example and save output
+- [ ] Add Safari 27 MCP when macOS node available
+- [ ] Connect to live GBP/VeriPath data pipelines
+## Success criteria
+One-click from dashboard: enter site + keyword → get site analysis report → get optimized content drafts → write results to vault.