The Agent Skills Directory

[PROMPT_INJECTION]: No attempts to override agent behavior or safety guidelines were detected. The skill uses instructional language consistent with professional security testing.
[DATA_EXFILTRATION]: The skill identifies and extracts sensitive information such as API keys and credentials from the target mini-program source code as part of a security audit. It does not attempt to access or exfiltrate the agent's own sensitive files or environment variables.
[COMMAND_EXECUTION]: The skill uses standard command-line tools (wxapkg, grep) for its intended purpose of unpacking and auditing code. All commands are relevant to the penetration testing workflow.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface as it processes untrusted data (the decompiled source code of mini-programs). However, the risk is negligible as it primarily uses static pattern matching (grep) and manual review rather than executing logic from the processed content.
Ingestion points: Analyzes decompiled .js, .wxml, and .json files.
Boundary markers: None explicitly defined in logic, but the methodology is focused on static analysis.
Capability inventory: Uses grep for auditing and wxapkg for decompression.
Sanitization: Not applicable as the skill performs pattern matching for security assessment purposes.
[SAFE]: The methodology includes explicit security boundaries, advising the user to only test authorized targets and avoid actual financial transactions during testing.

wechat-miniprogram-pentesting