Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
aaaaqwq avatar

Inference Optimizer

  • 11 installs
  • 82 repo stars
  • Updated August 2, 2026
  • aaaaqwq/claude-code-skills

inference-optimizer is an OpenClaw skill that audits token usage, purges stale sessions, and optimizes inference speed.

About

inference-optimizer is an OpenClaw skill that audits token usage and reduces inference cost and latency. On /optimize or /audit it runs an audit script that reports workspace sizes, session counts and next steps, and on approval it purges stale sessions and daily memory stubs. It exists because loaded workspace files and accumulated session history slow the model without an obvious cause. A developer uses it to keep an agent's context lean and fast.

  • Audits OpenClaw token usage and workspace/session footprint
  • Purges stale session files and daily memory stubs to speed up inference
  • Runs via /optimize and /audit commands with a purge step on approval

Inference Optimizer by the numbers

  • 11 all-time installs (skills.sh)
  • Ranked #2,174 of 3,280 Productivity & Planning skills by installs in the Skillselion catalog
  • Data as of Aug 3, 2026 (Skillselion catalog sync)
At a glance

inference-optimizer capabilities & compatibility

Capabilities
token optimization · session cleanup
Use cases
token optimization
Pricing
Free
From the docs

What inference-optimizer says it does

Audit OpenClaw token usage, purge stale sessions, and optimize inference speed.
SKILL.md
Instead of guessing where tokens go, you get workspace sizes, session counts, and actionable next steps.
README.md
npx skills add https://github.com/aaaaqwq/claude-code-skills --skill inference-optimizer

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs11
repo stars82
Last updatedAugust 2, 2026
Repositoryaaaaqwq/claude-code-skills

What it does

Use it to audit agent token usage and purge stale sessions so inference stays fast and cheap.

Who is it for?

Keeping an OpenClaw agent's context lean by auditing token usage and purging stale sessions.

Skip if: Non-OpenClaw runtimes or model-level inference tuning beyond workspace/session cleanup.

When should I use this skill?

The user sends /optimize or /audit, asks to purge sessions, or wants a token/workspace audit.

What you get

A token/workspace audit and a purge of stale sessions that restore inference speed.

  • A token/workspace audit report
  • Purged stale sessions and stub files

Files

SKILL.mdMarkdownGitHub ↗

Inference Optimizer

Optimize OpenClaw for maximum inference speed and minimum token usage. When user sends /optimize or /audit, exec the audit script and return raw output. When user approves purge, exec the purge script.

Triggers

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

Workflow

1. Audit: Exec bash <skill_dir>/scripts/openclaw-audit.sh, return raw output. Do NOT ask what to optimize. Do NOT list options. 2. Purge: If user approves purge after audit, exec bash <skill_dir>/scripts/purge-stale-sessions.sh, return raw output. 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.

Allowlist

Purge script needs: find, find *, find **, rm, rm *, rm **, bash, bash *, bash **. Add to exec-approvals.json for the agent if purge is blocked.

Related skills

FAQ

What does the audit report?

Workspace sizes, session counts and actionable next steps for reducing tokens.

When does it purge?

Only after the audit and once the user approves the purge step.

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.