Security audit
Security checks for vulnerabilities and agentic risk
This is a local encoding/decoding reference skill with some rough, potentially unsafe examples, but no evidence of hidden execution, exfiltration, persistence, or malicious behavior.
Before installing, note that this skill may lead an agent to run local shell commands and write or edit files. Use explicit paths, avoid running examples on important files without backups, and do not configure or share API keys unless a specific trusted workflow explains exactly why they are needed.
The display name and summaries are entirely Chinese, and the rest of the skill documentation is also presented in Chinese, with no indication that users may choose another language. Per the policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is clearly documented and justified.
The example writes decoded data directly to a local file, which can overwrite or create files without an explicit warning or safer handling guidance. In an agent context, users may paste untrusted filenames or content and trigger unintended filesystem modification.
The documentation recommends an in-place sed command that directly modifies files to remove a BOM, but it does not warn that the operation is destructive. In-place edits are risky in agent-driven workflows because they can silently alter user data or damage important files if applied to the wrong path.
The skill documents only local encoding/decoding and JWT parsing, yet later introduces external API usage and API key configuration without clearly tying that behavior to any feature. This inconsistency can mislead users into exposing secrets or enabling unnecessary networked behavior, expanding trust and data-exfiltration risk beyond the stated scope.
