insight-adversarial
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection attacks because it ingests and processes untrusted user data (such as plans, designs, or code) as the primary subject for its adversarial analysis passes.
- Ingestion points: Untrusted content enters the agent's context through the user's input target which is then processed through multiple critique phases.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands (e.g., XML tags or clear 'ignore instructions' warnings) are used when processing the target content in the prompt templates.
- Capability inventory: The skill has the ability to read local files (std:fs.read) and can recommend the execution of further skills such as 'plan-create' or 'plan-execute' based on the review synthesis.
- Sanitization: No sanitization, escaping, or validation of the input content is performed before it is analyzed by the critic personas.
Audit Metadata