The Agent Skills Directory

[DATA_EXFILTRATION]: The scripts check_sensitive_words.py, fetch_xhs_trends.py, and fetch_explosive_covers.py programmatically read sensitive shell configuration files such as ~/.zshrc, ~/.bashrc, and ~/.profile to locate API keys. Accessing these files is a high-risk activity as they often contain other sensitive credentials or environment variables.
[PERSISTENCE]: The SKILL.md and README.md instructions guide users to append export commands to their shell profile files (e.g., echo 'export REDFOX_API_KEY=ak_xxx' >> ~/.zshrc). This is a persistence mechanism that modifies the user's shell environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by crawling external URLs provided by the user via the check_sensitive_words.py script. The content is extracted using Playwright and BeautifulSoup and then processed by the AI agent without explicit sanitization, creating a surface for indirect prompt injection attacks.
Ingestion points: scripts/check_sensitive_words.py via the --url argument.
Boundary markers: None identified in the processing of remote content.
Capability inventory: Network access to redfox.hk, file writing to the local system, and shell command execution via subprocess.
Sanitization: None; raw text is extracted and passed to the agent.
[DYNAMIC_EXECUTION]: The scripts fetch_xhs_trends.py and fetch_explosive_covers.py utilize subprocess.run to execute PowerShell commands on Windows systems to retrieve user-level environment variables.
[COMMAND_EXECUTION]: The skill requires the execution of playwright install chromium as part of its setup process, which involves downloading and executing external binaries from the Playwright project.

xiaohongshu-creator