docs: rewrite CLAUDE.md per best practices, add BOOKMARKS.md

.claude/CLAUDE.md

@@ -1,13 +1,12 @@
 # GPS - Global Positioning System MCP Server
 
-Read-only caching tier for org and engineering data. Materializes people, teams, issues, features, releases, and governance into a SQLite database. Optimized for sub-millisecond agent queries over MCP — no auth, no rate limits.
+Read-only caching tier for org and engineering data. Materializes people, teams, issues, features, releases, and governance into a SQLite database and serves them over MCP.
 
 ## Structure
 - `mcp_server.py` — MCP server (stdio default, `--http` for HTTP)
 - `scripts/` — ETL pipeline (build_db.py, test.sh)
 - `data/` — source files + gps.db (user data gitignored, acme-* examples tracked)
 - `deploy/` — Kubernetes manifests + deploy.sh automation
-- `docs/` — ADRs, deployment, customization, schema reference
 - `governance/` — org policies (auto-loaded into DB)
 
 ## Commands
@@ -18,26 +17,11 @@ uv run mcp_server.py                         # start MCP server (stdio)
 uv run mcp_server.py --http                  # start MCP server (HTTP :8000)
 ```
 
-## MCP Tools (read-only, no auth)
-- `lookup_person` — find people by name, email, or user ID
-- `list_team_members` — list team roster with roles and components
-- `search_issues` — filter issues by status, priority, assignee, component, label, keyword
-- `get_feature_status` — feature details: progress, RICE score, releases, teams
-- `release_risk_summary` — flags features under 80% complete near milestones
-- `list_documents` — governance documents with table of contents
-- `get_document` — full governance document by ID
-- `get_document_section` — single section by fuzzy heading match
-- `get_gps_version` — version and build metadata
-
-## MCP Resources
-- `gps://schema` — full DDL with row counts (read this first)
-- `gps://catalog` — data source inventory
-
-## ACP Integration
-GPS runs as a sidecar MCP in every ACP pod. No auth needed — read-only org data. Wire via managed settings, init container, or HTTP sidecar. See docs/DEPLOYMENT.md.
-
 ## Rules
 - Never commit: csv, xlsx, pdf, db, .env, tokens, secrets (except data/acme-* examples)
 - Database (data/gps.db) is always opened read-only
 - Use uv (not pip). Use ruff (enforced by pre-commit).
 - `scripts/test.sh` is bash — run directly, not via `uv run`
+
+## More Info
+See [BOOKMARKS.md](BOOKMARKS.md) for deployment docs, schema reference, and ADRs.

BOOKMARKS.md

@@ -0,0 +1,29 @@
+# Bookmarks
+
+## Table of Contents
+- [Deployment](#deployment)
+- [Schema & Data](#schema--data)
+- [Architecture Decisions](#architecture-decisions)
+
+---
+
+## Deployment
+
+### [docs/DEPLOYMENT.md](docs/DEPLOYMENT.md)
+Deployment guide covering local dev, ACP integration, container builds, OpenShift/Kubernetes manifests, and CI/CD workflow.
+
+### [deploy/deploy.sh](deploy/deploy.sh)
+CLI for building images, pushing to registries, and applying kustomize manifests. Run `deploy/deploy.sh` with no args for usage.
+
+## Schema & Data
+
+### [docs/SCHEMA.md](docs/SCHEMA.md)
+Full database schema reference with table descriptions and column details.
+
+### [docs/CUSTOMIZATION.md](docs/CUSTOMIZATION.md)
+How to add custom data sources, configure ETL, and extend the database.
+
+## Architecture Decisions
+
+### [docs/adr/](docs/adr/)
+Architecture Decision Records for GPS design choices.