The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The 'Audit' mode in SKILL.md fetches content from user-provided URLs for analysis. This creates a surface for indirect prompt injection where a malicious website could attempt to influence the agent's behavior during the audit process.
Ingestion points: SKILL.md (Phase 1: Reconnaissance in Audit Mode) uses WebFetch to ingest untrusted data from external websites.
Boundary markers: The instructions do not define specific boundary markers or 'ignore' instructions for the fetched content.
Capability inventory: The skill uses WebFetch to retrieve data, Write to create reports, and Bash(open *) to execute system commands to view results.
Sanitization: No explicit sanitization or filtering of the fetched HTML content is mentioned before it is processed for copy analysis.
[COMMAND_EXECUTION]: The skill uses the open command via the Bash tool to automatically display generated HTML reports to the user. While intended for convenience, this involves executing system commands based on generated file paths.
[EXTERNAL_DOWNLOADS]: The skill documentation in README.md suggests using npx skills add for installation and git clone for manual setup. These are standard platform installation patterns but involve downloading external code.

language-market-fit