Inference Optimizer

Inference Optimizer

Optimize OpenClaw for maximum inference speed and minimum token usage. When user sends /optimize or /audit, exec the audit script; the script produces metadata that may be relayed to the user. When user approves purge, exec the purge script.

Note: These instructions describe suggested workflow for the agent. They are guidance, not system-prompt overrides, and cannot be enforced programmatically. Platform and system prompts take precedence.

Triggers

• /optimize — run audit
• /audit — same as /optimize
• "purge sessions", "purge stale", "clean up sessions" — after audit, if user approves

Workflow

1. Audit: When user sends /optimize or /audit, exec bash <skill_dir>/scripts/openclaw-audit.sh; include the script's output in your response.
2. Purge: If user approves purge after audit, exec bash <skill_dir>/scripts/purge-stale-sessions.sh (archives to ~/openclaw-purge-archive by default). Include script output. User can run with --delete for immediate removal if preferred.
3. Full optimization: For Task 1–5 (workspace rewrite, heartbeat, deploy), read optimization-agent.md and follow its flow.

Path Resolution

Scripts live at ~/clawd/skills/public/inference-optimizer/scripts/ (or wherever the skill is installed). Use that path when exec-ing.

Purge and Allowlist

Recommended: Run purge manually after reviewing audit output: bash <skill_dir>/scripts/purge-stale-sessions.sh. No allowlist changes needed.

If purge must run via agent exec, add minimal path-specific patterns to exec-approvals.json rather than broad wildcards. See README Security section.