ce-deepen-spec
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill processes user-provided specification files which may contain instructions intended to influence agent behavior.
- Ingestion points: Specification files read in SKILL.md (Phase 0).
- Boundary markers: Explicit instructions in SKILL.md mandate treating external text as untrusted input.
- Capability inventory: File system write operations (Phase 6) and research sub-agent calls (Phase 3).
- Sanitization: Constraints require redaction of PII and credentials.
- [SAFE]: Evaluation Suite. The references/evals.yaml file contains prompt injection test cases used for regression testing and is not part of the active operational instructions.
- [SAFE]: Research Grounding. The skill retrieves data from trusted organizations like OpenAI via research agents, which is categorized as safe external data retrieval.
Audit Metadata