The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites, which may contain instructions designed to manipulate the agent's behavior.
Ingestion points: The fetch_content function in scripts/monitor.py retrieves content from arbitrary user-provided URLs.
Boundary markers: Although the script uses markers like ─── and Diff # for output formatting, it does not explicitly instruct the agent to ignore instructions embedded within the fetched content.
Capability inventory: The script writes snapshots and alerts to the user's home directory (~/.web-monitor/) and performs network requests.
Sanitization: The script uses BeautifulSoup to strip HTML scripts and styles, but it does not sanitize the resulting text content against adversarial prompt payloads.
[EXTERNAL_DOWNLOADS]: The skill fetches data from remote servers as part of its core monitoring functionality, which involves downloading content from URLs provided by the user.
[COMMAND_EXECUTION]: The skill facilitates the execution of a local Python script (scripts/monitor.py) using the uv tool to manage web monitoring tasks.

web-monitor