deep-planning
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill enforces a read-only boundary, forbidding modifications to files or external systems and limiting the agent to static inspection or verification commands.
- [PROMPT_INJECTION]: The skill addresses indirect prompt injection through behavioral constraints. 1. Ingestion points: Repository files (AGENTS.md, ADRs, code, tests), logs, and captured webpages. 2. Boundary markers: Explicit instructions to treat analyzed content as evidence rather than instructions and forbidding scope expansion. 3. Capability inventory: Authorization to run tests, builds, and diagnostics for verification purposes. 4. Sanitization: Relies on explicit directives to disregard instruction-like patterns within ingested data.
Audit Metadata