The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local Python script scripts/wechat_similar_accountr.py to process account analytics. The script utilizes standard modules to query parameters provided by the user.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to the vendor domain redfox.hk to obtain account metadata and article lists. This is a legitimate functional requirement for the skill.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the processing of externally sourced content.
- Ingestion points: Data entering the context includes account names and article titles/summaries fetched from the RedFox API in scripts/wechat_similar_accountr.py.
- Boundary markers: External data is presented in markdown tables without explicit delimiters or warnings to the agent.
- Capability inventory: The skill can perform network operations and execute local commands via the Python script.
- Sanitization: Content from the API is parsed as JSON but is not specifically filtered or escaped before being rendered in the agent's output.
[DATA_EXFILTRATION]: The provided Python script contains functions to scan user shell configuration files (such as .zshrc and .bashrc) and the Windows Registry to locate the REDFOX_API_KEY. While this is done to facilitate authentication with the vendor's own API, it involves reading files that commonly contain sensitive user configuration.

wechat-similar-account