The Agent Skills Directory

[DATA_EXFILTRATION]: The Python script scripts/wechat_analyzer.py contains logic that accesses sensitive environment configuration files on the user's system to retrieve credentials.
Sensitive File Access: The function _read_from_shell_config attempts to read ~/.zshrc, ~/.bashrc, ~/.bash_profile, and ~/.profile.
Pattern Matching: It searches these files for the string export REDFOX_API_KEY=, potentially exposing other environment variables or sensitive configuration data stored in these files.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to the ingestion and processing of untrusted data from an external API.
Ingestion Points: Data enters the agent's context through the redfox.hk API, including article titles, descriptions, and account metadata (references/core_workflow.md).
Boundary Markers: No explicit delimiters or instructions are provided to the agent to distinguish between administrative instructions and content retrieved from the API.
Capability Inventory: The agent is instructed to execute a Python script (scripts/wechat_analyzer.py) and perform file system operations in the output/ directory.
Sanitization: There is no logic specified to sanitize or escape data retrieved from the external source before it is interpolated into the final markdown reports.
[COMMAND_EXECUTION]: The core functionality of the skill relies on the execution of a local Python script by the agent.
Subprocess Call: Instructions in references/core_workflow.md direct the agent to run python scripts/wechat_analyzer.py query --account_names [Name]. This capability allows for direct interaction with the host operating system.

wechat-account-analyzer