openclaw-optimizer_skill

Overview

This skill optimizes and troubleshoots OpenClaw (v2026.2.23+) workspaces with a CLI-first, advisory workflow. I audit cost, model routing, provider configuration, context health, cron automation, and agent identity/personality, then produce a prioritized plan with exact CLI commands, config patches, and rollback steps. Changes are suggested and applied only after explicit approval.

How this skill works

I run a safe, read-only audit of OpenClaw runtime and config, check provider and model status, validate routing and cron setups, and scan identity/personality files for conflicts and bloat. I produce prioritized recommendations, concrete CLI commands, and precise config patches, plus rollback commands and impact notes. If you request fixes, I present each change as a one-line CLI or patch and require explicit approval before applying.

When to use it

• You want to cut model costs with tiered routing or fallback chains.
• You need to add, validate, or remove a provider (API key or OAuth) safely.
• You’re troubleshooting startup failures, channel errors, or provider auth issues.
• You want cron jobs that run silently or need isolated NO_REPLY behavior.
• You need an audit and cleanup of agent identity/personality files (SOUL/IDENTITY/AGENTS/USER).
• You’re designing sub-agent architectures, context discipline, or multi-agent routing.

Best practices

• Run a full audit first (read-only) and review the prioritized plan before applying changes.
• Prefer CLI rolling .bak and nightly GitHub backups; create manual dated backups only for multi-file edits or upgrades.
• Use tiered routing: cheap for heartbeats, mid for daily tasks, premium for code/reasoning.
• Keep OAuth providers separate from API-key providers; prefer API keys for fragile providers (e.g., Google Gemini).
• Never switch models mid-conversation; change at session boundaries to preserve caches.
• Require explicit approval before any config mutation; always show a rollback command.

Example use cases

• Full audit: prioritized cost/reliability plan, exact CLI commands, and rollback — no changes applied.
• Add provider: walk-through CLI auth, models set, status probe, and restart steps for gateway.
• Model routing optimization: propose tiered routing config with exact models and fallbacks, plus rollback.
• Troubleshoot: diagnose an error message and provide the minimal fix and restart commands.
• Agent identity audit: detect conflicting prompts, reduce bloat, and supply patched personality files and restore steps.