Back to skill
Skillv1.0.0
ClawScan security
H test · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 16, 2026, 8:07 AM
- Verdict
- Benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's code, runtime instructions, and required environment variables are coherent with an Ezviz-based restaurant inspection integration; no unrelated credentials, hidden endpoints, or dangerous install steps were found.
- Guidance
- This skill appears to be what it says: it calls Ezviz APIs to fetch images and an Ezviz AI endpoint to analyze them. Before installing/using it: 1) Use least-privilege appKey/appSecret and rotate them regularly. 2) Be aware the script sends camera images and metadata to Ezviz endpoints — ensure you have legal/privacy consent. 3) Consider removing or reducing debug prints (response headers and raw JSON snippets) before running in production to avoid logging sensitive API responses. 4) The SKILL.md recommends a ≥4s interval between capture calls but the script uses a 1s delay; if you will run against many devices, increase the delay to avoid rate-limiting. 5) Test in a controlled environment first and audit network calls if you need higher assurance.
Review Dimensions
- Purpose & Capability
- okName/description, required env vars (EZVIZ_APP_KEY, EZVIZ_APP_SECRET, EZVIZ_DEVICE_SERIAL, EZVIZ_AGENT_ID), declared endpoints (open.ys7.com, aidialoggw.ys7.com), and code (uses requests to call those APIs) are all consistent with an Ezviz camera + AI analysis inspection skill.
- Instruction Scope
- noteSKILL.md and the script stay within the stated scope (obtain token, capture image, call AI analysis). Minor inconsistencies: the README warns about a 4‑second capture interval but the script only delays 1 second between devices (risk of hitting rate limits). The doc claims tokens are not written to logs/disk — the script does not persist tokens to disk, but it does print debug output (response headers and portions of raw JSON responses). This may expose partial API responses in console logs; consider removing or reducing debug prints in production.
- Install Mechanism
- okNo install spec beyond a pip dependency on 'requests' (declared in metadata). The script only uses standard libs + requests, so the install footprint is minimal and proportional.
- Credentials
- okAll required environment variables are directly relevant to Ezviz API access and the AI agent (app key/secret, device serial(s), agent id). No unrelated credentials or broad system config paths are requested.
- Persistence & Privilege
- okThe skill does not request persistent 'always' inclusion and does not modify other skills or system-wide settings. It does not write tokens to disk according to the code. Autonomous invocation is allowed (platform default) and appropriate for this integration.