The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (clean_html.py and extract_metadata.py) using the uv run command to process HTML files. This is standard functionality for the skill's purpose.
[DATA_EXFILTRATION]: The skill reads local HTML files and writes extracted structured data to local JSON files. It does not perform any network operations, and no data exfiltration to external domains was detected.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its ingestion of untrusted HTML data.
Ingestion points: Local HTML files (referenced as PAGE.html) provided in the $ARGUMENTS input.
Boundary markers: The instructions lack explicit delimiting or prompt-level directives to ignore instructions that might be embedded within the HTML content.
Capability inventory: The skill utilizes Bash, Read, and Write tools, and the Python scripts have access to the local file system.
Sanitization: The clean_html.py script provides basic sanitization by stripping <script>, <style>, and <svg> tags, as well as HTML comments, which mitigates some technical attack vectors but does not prevent natural language instructions in the remaining text from influencing the model.

scrape-analyze-page