Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace (cost, model routing, context discipline, delegation, r...
Security Analysis
high confidenceThe skill is an instruction-only advisory playbook for auditing and proposing OpenClaw optimizations; its requested footprint (no installs, no env vars, no binaries) matches its stated purpose.
Name/description claim an advisory optimization playbook; the package is instruction-only, contains runbooks and examples, and does not request unrelated resources or credentials. The guidance (model routing, cron/heartbeat patterns, file paths like memory/) is coherent with an OpenClaw ops optimization skill.
SKILL.md and references instruct the agent to inspect workspace artifacts (memory/, references/, logs) and to produce exact config patches and rollback plans. It also suggests running audit-only commands (e.g., context-clean-up) and checking ClawHub auth. These actions are appropriate for an optimization advisor but will touch local workspace files — the skill explicitly requires explicit user approval before any persistent changes.
No install spec or code files beyond docs; README suggests optional ClawHub install as a convenience. No downloads, extracts, or third-party package installs are performed by the skill itself.
The skill declares no required environment variables, credentials, or config paths. Occasional optional checks (e.g., `npx clawhub whoami`, recommending openclaw-mem) are advisory and not required; nothing asks for unrelated secrets.
always:false and no code that auto-writes system-wide settings are present. The instructions explicitly mandate user approval before creating/updating cron jobs or applying persistent config patches. Autonomous invocation (model invocation enabled) is the platform default and not in itself a problem here.
Guidance
This skill is an advisory runbook — it contains detailed checks and exact patch suggestions but does not perform changes without your approval. Before installing/using it: (1) review any proposed config patches and rollback steps before accepting them, (2) be aware it will examine and recommend edits to workspace files (memory/, references/, cron definitions), so run audits or dry-runs first, (3) optional recommendations point to external tools/repos (ClawHub, openclaw-mem on GitHub); only follow those if you trust the source, and (4) no credentials are required by the skill itself, but if you choose to run suggested commands that call external services, ensure you only expose necessary tokens and understand what those tools do.
Latest Release
v1.2.0
Polish GTM + portability: clearer advisory contract, add license, tighten wording, and improve quick-start copy.
More by @phenomenoner
Published by @phenomenoner on ClawHub
