compact-for-task
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data through the
$ARGUMENTSvariable and interpolates it directly into the agent's instructions. - Ingestion points: The
$ARGUMENTSvariable inSKILL.mdreceives input from the user/next task. - Boundary markers: The skill lacks delimiters (e.g., XML tags, triple quotes) or specific instructions to the agent to treat the interpolated text as data rather than instructions.
- Capability inventory: The skill itself does not invoke tools, but it generates text that instructs the agent's future behavior (the
/compactcommand). - Sanitization: No escaping or validation is performed on the input before interpolation.
Audit Metadata