recoup-internal-consulting-copy-reviewer
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a file target and reads its content into the prompt context of a subagent. Untrusted data from this file could contain instructions designed to influence the agent's behavior. \n- Ingestion points: The target draft file defined in
{DRAFT_PATH}is read by the subagent in Step 3. \n- Boundary markers: The skill fails to use explicit delimiters or "ignore embedded instructions" warnings when interpolating the draft content into theReviewer briefprompt. \n- Capability inventory: The subagent is restricted to a read-only state, preventing it from modifying files directly, though the main agent is instructed to rewrite files based on the resulting notes. The main agent can perform file write operations. \n- Sanitization: No sanitization, escaping, or validation of the external file content is described in the workflow instructions.
Audit Metadata