The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/check_sensitive_words.py reads sensitive system files including ~/.bashrc, ~/.zshrc, ~/.bash_profile, and ~/.profile to locate the REDFOX_API_KEY credential.
[DATA_EXFILTRATION]: User data, including pasted text, file contents, and extracted webpage content, is sent to the external service redfox.hk for analysis.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @alipay/agent-payment package via npx and browser binaries via playwright install chromium.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external URLs and user-uploaded files (TXT, DOC, DOCX, etc.), creating a potential surface for indirect prompt injection attacks.
Ingestion points: External URLs via extract_from_web and uploaded files via extract_from_file in scripts/check_sensitive_words.py.
Boundary markers: None identified in the content processing or prompt assembly logic.
Capability inventory: The skill has the ability to perform network requests, file system read/write, and execute shell commands via subprocesses.
Sanitization: The script performs basic content length checks but does not implement sanitization or filtering to prevent prompt injection from external sources.

wechat-prohibited-word