Back to skill
Skillv1.0.0
ClawScan security
Model Cost Advisor · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignApr 26, 2026, 7:29 AM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's code and instructions are consistent with its stated purpose (estimating token usage and comparing live model prices); it fetches a community price DB from GitHub and caches it under ~/.hermes, and does not request secrets or unrelated system access.
- Guidance
- This skill appears to do what it claims: it downloads a community-maintained pricing JSON from GitHub and caches it in ~/.hermes, then analyzes task text locally to recommend models. Before installing/run: (1) verify you trust the litellm source (the script fetches raw.githubusercontent.com/BerriAI/litellm/...); (2) review the cached file (~/.hermes/model_pricing.json) if you want to inspect the data before use; (3) if you prefer no persistent files, run the scripts in a disposable environment or delete the ~/.hermes directory after use; (4) be aware it may optionally read HERMES_CURRENT_MODEL for comparison — set that only if you intend to share your current model. No credentials are required by the skill.
Review Dimensions
- Purpose & Capability
- okName/description match the implementation: the two scripts fetch pricing from litellm's JSON and perform task analysis, token estimation, and cost computation. There are no unrelated credentials, binaries, or services requested.
- Instruction Scope
- okSKILL.md instructs only to run fetch_pricing.py and advise.py with a task string; the runtime instructions do not direct the agent to read or exfiltrate unrelated files or environment variables. The only optional integration is comparing against HERMES_CURRENT_MODEL (an expected, optional env var).
- Install Mechanism
- noteNo install spec; scripts are instruction-only. fetch_pricing.py downloads a JSON from raw.githubusercontent.com (a well-known release host) — reasonable for live pricing. It writes a cache file to ~/.hermes/model_pricing.json, which is expected behavior but worth noting because it creates files in the user's home directory.
- Credentials
- okThe skill declares no required environment variables or credentials. The only env referenced in docs is HERMES_CURRENT_MODEL for optional comparison; no secrets (API keys, tokens) are requested or used by the provided code.
- Persistence & Privilege
- notealways:false (normal). The skill persists a cache to ~/.hermes/model_pricing.json and creates the directory if needed — reasonable for caching but it does create a persistent file in the home directory. It does not alter other skills or system-wide agent settings.