OCC | Slash Command & Skill Development

The Case Study: /ops Command

The /ops command was envisioned as a high-speed, LLM-free diagnostic tool for the entire VTS and OCC ecosystem. It allows for deterministic reporting of service health without incurring token costs or reasoning delays.

                // Logic Location: /app/skills/vts-status/scripts/check_status.js

                // Mechanism: Direct process table audit via ps aux

The Post-Mortem (Crash-Loop Incident)

Error: commands.native: Invalid input

An attempt was made to patch the global commands.native flag in openclaw.json with an object containing custom command logic.

Root Cause: The Zod validator for native only accepts true/false/"auto".
Result: Gateway validation failure and service downtime.
Correction: Custom command registry must live in commands.customCommands array.

Best Practices for Future Expansion

1. Logic-First

Decouple core logic from the LLM. Write standalone scripts in a skill's scripts/ folder for speed and reliability.

2. Schema Safety

Always verify the openclaw.json structure via config.get before patching. Never assume object structures.

3. Determinism

Report absolute truths (PIDs, Ports, Status) via system calls. Use the LLM only for summarizing when necessary.

4. Standardized Skills

Follow the SKILL.md + scripts/ + references/ anatomy for cross-agent compatibility.