trace
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill utilizes the
$ARGUMENTSvariable inSKILL.mdto ingest user questions, which then dictate the agent's codebase analysis path. The lack of explicit boundary markers or instructions for the agent to disregard potential commands inside the user input creates a surface for indirect prompt injection. - Ingestion points: User questions are ingested via the
$ARGUMENTSvariable in theSKILL.mdfile. - Boundary markers: The skill does not define specific delimiters or instructional boundaries to separate untrusted user input from the agent's core operational logic.
- Capability inventory: The agent performs file searches and read operations across the project codebase to identify logic triggers and code hops.
- Sanitization: There is no evidence of sanitization, filtering, or validation performed on the raw user input before it is used to drive agent behavior.
Audit Metadata