best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze untrusted data from external sources during the audit process.
- Ingestion points: Data enters the agent's context through
lighthouse_auditresults, browser console logs, and network failure reports, as described in the audit workflow inSKILL.md. - Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings for separating tool output from instruction context.
- Capability inventory: The agent is expected to use file system tools and MCP tools to analyze and fix code based on these potentially untrusted inputs.
- Sanitization: Although the skill promotes sanitization for the target application (e.g., via
DOMPurify), it does not explicitly describe sanitization or validation protocols for the agent's own processing of external audit data.
Audit Metadata