The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script scripts/conversion_audit.py to perform automated analysis of HTML pages. The script is executed with user-provided file paths or URLs.
[EXTERNAL_DOWNLOADS]: The conversion_audit.py script uses urllib.request.urlopen to download content from arbitrary URLs provided via the CLI argument --url.
[DATA_EXFILTRATION]: The ability to fetch content from any destination URL provides a vector for outbound network requests (SSRF). While no explicit exfiltration of sensitive data was found, the script can be directed to internal or external network resources by the agent or a user.
[PROMPT_INJECTION]: The skill is designed to ingest and analyze external marketing pages, creating a surface for Indirect Prompt Injection. Malicious instructions embedded in analyzed HTML (e.g., in metadata, alt text, or hidden tags) could attempt to influence the agent's behavior.
Ingestion points: External URLs fetched via the audit script and local marketing page files read by the agent.
Boundary markers: Absent; the instructions do not provide delimiters or warnings to ignore embedded instructions in the analyzed data.
Capability inventory: Execution of Python scripts and unrestricted outbound network requests to user-provided URLs.
Sanitization: The audit script counts signals but does not sanitize or filter the content before the agent evaluates the page context.

page-cro