The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the command line to perform public scans using a specific vendor tool.
Evidence: npx @ora-ai/ax@0.5 audit https://<site> --json in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill fetches and executes a package from a public registry during the optional public scan phase.
Evidence: The skill instructs the agent to run an npx command which downloads the @ora-ai/ax package.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading repository files and crawling a running website, which could contain malicious instructions designed to influence the agent's report.
Ingestion points: The skill reads project files (README, manifests, route definitions) and observes running site content (HTML, forms, console errors) as described in SKILL.md and references/checks.md.
Boundary markers: No specific delimiters or safety instructions are mentioned for handling the content read from the repository or site.
Capability inventory: The skill has the ability to execute shell commands (npx), write files (webmcp-audit.md), and perform network requests (GET requests to the vendor API).
Sanitization: There is no evidence of specific sanitization or filtering of the content ingested from the audited website or repository before it is processed for the report.

audit