Cross-Channel & Cross-Agent Skill Sharing#

Execute this workflow whenever workspace capability changes#

1. Detect capability change source:
   • New package/tool installed in workspace
   • Existing tool upgraded and behavior changed
   • New env/interpreter path required for capability
2. Update TOOLS.md with operational facts:
   • Exact runtime/interpreter path
   • Install command
   • Verify command
   • Package versions
   • One-line usage note
3. Create or update a reusable skill when pattern is repeatable:
   • Add/adjust SKILL.md
   • Add scripts under scripts/ for deterministic steps
   • Keep instructions concise and invocation-focused
   • Prefer adding a verification script for fast capability checks
4. Log event in memory/YYYY-MM-DD.md:
   • What changed
   • How to verify
   • Any limits/known caveats
5. Inform current conversation:
   • Summarize new capability
   • Show exact invocation command or trigger phrase
6. Propagate to active related sessions/agents:
   • If likely relevant, send brief sync via sessions_send
   • Include only invocation + critical caveat

Required output format in chat (short)#

• New capability:
• Use it with: <exact command / trigger>
• Scope: Applies to all channels and sessions using this workspace
• Caveat:

Quick Verification Helper#

Use bundled script to verify interpreter + modules before claiming capability:

bash

scripts/check_capability.sh <python-bin> <module1> [module2 ...]

Example:

bash

scripts/check_capability.sh ~/.openclaw/workspace/.venv-img/bin/python cv2 PIL numpy

Guardrails#

• Do not assume global python/node paths; prefer workspace-scoped env paths
• Do not claim capability before running verify command
• Keep cross-session notices minimal to avoid spam
• If capability is channel-specific (provider restriction), state that explicitly