coding-agent-docs-optimization
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data by design, as its primary function is to fetch and analyze third-party documentation provided by the user.
- Ingestion points:
SKILL.mdandreferences/agent-surfaces.mdcontain instructions for the agent to fetch documentation pages, markdown entry points, and machine-readable specs from the web. - Boundary markers: The instructions do not specify the use of delimiters to separate ingested documentation from the agent's core instructions, as the content is treated as audit data.
- Capability inventory: The skill utilizes network access to retrieve documentation and involves performing 'cold runs' to test integration snippets found in the documentation.
- Sanitization: No specific sanitization of external documentation content is performed before the agent audits it.
- [EXTERNAL_DOWNLOADS]: The skill performs outbound HTTP requests as part of its documented workflow to verify that documentation surfaces are not blocked by CDNs or WAFs.
- Evidence:
SKILL.md(Workflow Step 1) andreferences/agent-surfaces.md(Access Check) explicitly instruct the agent to fetch documentation pages and record HTTP status codes to ensure agents can access the content.
Audit Metadata