Analysis: Claude Planning Doc vs Current Setup

Conflicts & Differences

1. Agent ID: "ada" vs "main" ✅ COMPLETED

• Status: Renamed from main → ada on 2026-03-01
• Workspace: Moved to /home/sean/.openclaw/workspace-ada for consistency
• Memory refs: Updated across all files
• Result: Full rebrand complete. Agent ID, workspace, and memory all aligned.

2. Model Strategy

• Plan: Sonnet 4-5 for most, Opus 4-5 for K2
• Current: Gemini 2.5 Flash for all (speed priority), Opus for subagents
• Decision needed: Sean explicitly asked for speed-first with Opus as subagents. The plan's model choices are more expensive. Our current setup aligns with Sean's verbal instructions, which supersede the plan.

3. Env Var Naming Convention

• Plan: TELEGRAM_BOT_TOKEN_ADA, DISCORD_BOT_TOKEN_K2 (suffix style)
• Current: K2_TELEGRAM_TOKEN, CORA_DISCORD_TOKEN (prefix style)
• Impact: Cosmetic only — both work. Suffix style is arguably more organized.
• Recommendation: Could standardize later, but not worth disrupting a working setup.

4. Agent-to-Agent Communication (IMPORTANT)

• Plan: Synergy can ONLY talk to Winston. Winston is the bridge.
• Current: All 5 agents in the allow list (tools.agentToAgent.allow)
• Action needed: Restrict per-agent subagents.allowAgents to match the plan.

5. Config Structure: $include vs Monolithic

• Plan: Split into config/*.json5 files with $include
• Current: Single openclaw.json
• Recommendation: The $include approach is cleaner for 5 agents but adds complexity. Can migrate later.

6. NAS-based Secrets with build-env.sh

• Plan: Secrets on NAS, assembled via build-env.sh
• Current: Secrets directly in ~/.openclaw/.env
• Recommendation: Good security practice to source from NAS. Can implement build-env.sh when NAS mount is confirmed.

7. Discord for Cora & Synergy

• Plan bindings: Only ada, k2, winston on Discord (Cora and Synergy missing from Discord bindings)
• Current: All 5 agents have Discord bots
• Note: Plan's channels.json5 also omits Cora and Synergy from Discord accounts. This may have been intentional or an oversight.

8. IDENTITY.md vs SOUL.md

• Plan: Uses IDENTITY.md for personality
• Current: Uses SOUL.md (OpenClaw convention)
• Recommendation: Keep SOUL.md as OpenClaw expects it. Can add IDENTITY.md as supplementary.

9. Logging

• Plan: redactSensitive: "tools" (redact only in tool output)
• Current: redactSensitive: "on" (redact everywhere)
• Recommendation: Keep "on" — more secure.

10. Session Config

• Plan: scope: "per-sender", daily reset at 4am, maxPingPongTurns: 3
• Current: Default session config
• Recommendation: Worth implementing the ping-pong limit to prevent agent loops.

Items Already Completed (from Plan's Checklist)

• [x] Register 5 Telegram bots
• [x] Create Discord applications for all 5
• [x] Get Sean's Telegram user ID (8494006989)
• [x] Get Sean's Discord user ID (780554363768012820)
• [x] agents add for all 5
• [x] Seed workspace files
• [x] openclaw doctor validated
• [x] Gateway running
• [x] All agents paired on Telegram and Discord

Still TODO

• [ ] Select ElevenLabs voice IDs for each agent
• [ ] Implement per-agent subagents.allowAgents restrictions
• [ ] Consider NAS-based secret management (build-env.sh)
• [ ] Consider $include config split (future)
• [ ] Configure Synergy with wife's input