The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a standalone Python script, scripts/diagnose.py, which is intended to be executed to perform network diagnostics and HTML analysis.
[DATA_EXFILTRATION]: The skill performs network operations to external, non-whitelisted domains via urllib.request in scripts/diagnose.py. While these operations are necessary for the skill's diagnostic purpose, they involve outbound connections to arbitrary hosts. No exfiltration of local sensitive data was observed.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it retrieves and processes HTML content from external sites, which is then presented to the agent.
Ingestion points: The fetch function in scripts/diagnose.py retrieves content from arbitrary user-provided URLs.
Boundary markers: The diagnostic results are presented in a structured report format, but the skill does not include explicit boundary markers or instructions to the agent to treat the ingested HTML content as untrusted.
Capability inventory: The script uses urllib.request for network access and html.parser.HTMLParser for content extraction.
Sanitization: The script performs basic HTML parsing and text extraction but does not sanitize the content to prevent the inclusion of malicious instructions intended for the agent.

fix-broken-scraper