The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The script scripts/check_sensitive_words.py scans local shell configuration files including ~/.bashrc, ~/.zshrc, .bash_profile, and .profile to retrieve the REDFOX_API_KEY credential.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from arbitrary URLs and user-uploaded files, which presents a surface for indirect prompt injection.\n
Ingestion points: Untrusted data enters the agent context through the extract_from_file and extract_from_web functions in scripts/check_sensitive_words.py.\n
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the ingested content as untrusted data.\n
Capability inventory: The skill possesses the ability to perform network requests via requests and playwright, and can write files to the local directory (e.g., ./公众号_优化文案_*.txt).\n
Sanitization: No validation or sanitization is performed on the ingested content before it is processed or sent to the API.\n- [EXTERNAL_DOWNLOADS]: The skill downloads content from user-specified URLs using the requests library and the playwright browser engine.

wechat-prohibited-word