Back to skill
Skillv1.1.0
ClawScan security
OpenClaw Cost Optimizer · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignFeb 14, 2026, 6:31 PM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's requirements and instructions align with its stated purpose: it needs the OpenClaw CLI and an OpenRouter API key to add model aliases and log savings; nothing requested appears unrelated or excessive.
- Guidance
- This skill appears to do what it says: add model aliases via the OpenClaw CLI and log savings in ~/.openclaw/cost-tracker.md. Before installing: 1) Verify the openclaw binary on your system is from a trusted source; the skill relies on that CLI to change your model aliases. 2) Keep your OPENROUTER_API_KEY limited in scope if possible; the skill will use it to configure models. 3) When the agent presents exec approvals, inspect the exact commands it will run before approving — the SKILL.md enforces that the agent run commands automatically. 4) Back up your existing OpenClaw/OpenRouter configuration so you can revert if needed. 5) If you want extra caution, test the process in a non-production/staging environment first. The skill's strict verbatim prompts and exec-first behavior are unusual but not inherently malicious; they just increase the need for the user to review each automated action.
Review Dimensions
- Purpose & Capability
- okName/description match the declared requirements: the skill manipulates OpenClaw/OpenRouter model aliases, so requiring the openclaw binary and OPENROUTER_API_KEY is coherent. No unrelated service credentials or unusual binaries are requested.
- Instruction Scope
- noteSKILL.md is prescriptive: it insists on exact verbatim prompts, instructs the agent to run all setup via the exec tool, and says it will create ~/.openclaw/cost-tracker.md. This is consistent with an automation skill, but the exact-text requirements and insistence on agent-run execs mean users should review every exec approval carefully.
- Install Mechanism
- okInstruction-only skill with no install spec or downloads. Nothing is written to disk by an installer; the only persistent artifact is the tracking file the skill creates under ~/.openclaw, which is declared in metadata.
- Credentials
- okOnly OPENROUTER_API_KEY is required (declared). That credential is necessary to configure models through OpenRouter/OpenClaw; no additional unrelated secrets or config paths are requested.
- Persistence & Privilege
- noteThe skill will create ~/.openclaw/cost-tracker.md to log activity and relies on exec to run commands. always:false and normal autonomous invocation are used. This level of persistence and autonomous command-execution is reasonable for a setup/automation skill but increases the importance of reviewing each exec approval.