The Agent Skills Directory

[DATA_EXPOSURE]: The script scripts/check_sensitive_words.py implements a routine _get_api_key() that scans several sensitive local shell configuration files (~/.zshrc, ~/.bashrc, ~/.bash_profile, ~/.profile, ~/.zprofile) to locate the REDFOX_API_KEY. Accessing these files can expose other environment variables or secrets stored in the user's shell profiles.
[DATA_EXFILTRATION]: The skill sends text content provided by the user (or extracted from files and URLs) to the external endpoint https://redfox.hk/story/api/cozeSkill/sensitiveWordSearch for processing. While this is the stated purpose of the skill, it involves the transmission of potentially sensitive user data to a third-party domain.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the playwright library and the execution of playwright install chromium, which downloads and installs the Chromium browser binary into the execution environment.
[COMMAND_EXECUTION]: The core workflow instructs the agent to create and write optimization results to local text files (e.g., ./小红书_优化文案_*.txt) using file system operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad ingestion surface for untrusted data, including local files (DOCX, TXT, CSV, MD) and remote web pages via the extract_from_web function.
Ingestion points: extract_from_file and extract_from_web in scripts/check_sensitive_words.py.
Boundary markers: No explicit instruction delimiters or boundary markers are used when processing the extracted text.
Capability inventory: The agent can write files to the local disk and perform network requests to the vendor's API.
Sanitization: The script uses BeautifulSoup to strip HTML tags from web content, but the extracted plain text is passed directly to the detection logic without further instruction-stripping filters.

xiaohongshu-prohibited-word