AndonQ

This skill appears to be a real Tencent Cloud Andon ticket + SmartQA client, but there are a few red flags you should consider before installing or using it: - Credentials disclosure: The skill needs TENCENTCLOUD_SECRET_ID and TENCENTCLOUD_SECRET_KEY to query tickets. The SKILL.md tells you to write them permanently into your shell profile (~/.bashrc / ~/.zshrc or Windows user env). Persisting AK/SK in startup files increases exposure — prefer using ephemeral environment variables, a dedicated least-privilege key, or a short-lived credential mechanism if available. Rotate the keys after initial use. - Metadata mismatch: The registry metadata lists no required env vars while the skill and docs require AK/SK; this mismatch can be accidental but it hides important security context from users. Treat the skill as requiring sensitive credentials until proven otherwise. - Version check subprocess: The environment checker runs 'clawhub inspect <slug>' via subprocess. That call will fail harmlessly if clawhub is not installed, but it means the script assumes an external CLI may exist; review what that CLI would do in your environment before installing it. - Code review & network behavior: The Python code included is readable and shows HTTP(S) requests to Tencent endpoints (tandon.tencentcloudapi.com, andon.cloud.tencent.com, cloud.tencent.com). There is no obvious credential exfiltration in these files, but you should run the code locally or review it fully before supplying credentials. Practical steps: - If you want to proceed: create a dedicated least-privilege Tencent Cloud API key (scope-limited if possible), do not store it permanently in shell startup files — set env vars only for the session or use a secrets manager, and rotate the key afterwards. - Run check_env.py locally (it is advertised read-only) to validate behavior before invoking ticket APIs. - Inspect/grep the code locally for any unexpected outbound hosts beyond the documented Tencent hosts. - If you cannot audit the code, avoid entering long-lived credentials into persistent configs; consider running this skill in an isolated environment or container. Confidence: medium — the functionality and network targets match the advertised purpose, but the undeclared credential requirement and the instruction to persist secrets into shell startup files are disproportionate and merit caution.