spec-interview
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTION
Full Analysis
- PROMPT_INJECTION (HIGH): The skill is susceptible to Indirect Prompt Injection (Category 8) due to its core functionality of processing external data.
- Ingestion points: Reads content from
plan.mdor user-specified technical documents (File: SKILL.md). - Boundary markers: Absent. The instructions do not provide delimiters or specific commands to ignore instructions embedded within the ingested data.
- Capability inventory: Uses
AskUserQuestionToolfor interactive dialogue and has the authority to write output back to the filesystem (File: SKILL.md). - Sanitization: None detected. There is no mechanism to prevent malicious instructions inside a
plan.mdfrom overriding the agent's intended role. - Risk: An attacker-controlled document could contain instructions that cause the agent to exfiltrate data through its questions or write malicious code to the local environment under the guise of a 'refined specification'.
Recommendations
- AI detected serious security threats
Audit Metadata