quill: add sovereign AI quick-reference draft

wizard-work/quill/sovereign-ai-quickref.md

@@ -0,0 +1,55 @@
+SOVEREIGN AI STACK — QUICK REFERENCE
+=====================================
+Operator: Alexander
+Last updated: 2026-11-02
+
+HERMES PROFILES
+---------------
+- Profile storage: /root/.hermes/profiles/
+- Active profile: quill
+- Per-profile: skills/, plugins/, cron/, memories/
+- NEVER edit another profile's files without explicit direction.
+
+SKILLS
+------
+- Load with: skill_view(name='<skill>')
+- Available categories: autonomous-ai-agents, creative, email, media, note-taking, productivity, research, social-media, software-development, web
+- Authoring: hermes-agent-skill-authoring skill
+
+CONNECTORS
+----------
+- Managed via: manage_connections (action=status/connect/reconnect)
+- Hosted: slugs like gmail, notion, linear
+- MCP servers: {"name": "<name>", "mcp": true}
+- Never hand-edit mcp_servers config.
+
+FREE-TIER RULES
+---------------
+- Zero paid vendors. Free credits only.
+- Verify every result; "executor said done" is NOT a test passed.
+
+WIZARD LANES
+------------
+- Draft: fast writing and drafts (this wizard)
+- Other lanes: inspect, build, debug, research, deploy
+- Each wizard ships artifacts, not plans.
+
+TROUPE REPO
+-----------
+- Path: /root/troupe-standup/
+- Wizard work: wizard-work/<wizard-name>/
+- Commit identity: -c user.name=<wizard> -c user.email=<wizard>@agents.invalid
+- Push token: $FORGE_TOKEN (never printed)
+
+EVIDENCE LAW
+------------
+- Prepared != observed.
+- Mark measured vs claimed.
+- Fabricated output is fraud.
+
+STANDARD COMMAND PATTERNS
+-------------------------
+- Batched reads: parallel tool calls, no serialization unless dependent.
+- Code execution: execute_code for 3+ tool calls with logic.
+- File edits: patch, write_file (read first for existing files).
+- Terminal: avoid cat/grep/find/sed/awk — use the dedicated tools.