blueprint-research
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Dynamic Execution] (LOW): The skill uses 'Skill(skill: skill)' in Step 5 to dynamically load logic based on the 'coding_styles' research output in SKILL.md. This pattern of loading from computed paths is downgraded to LOW as it is central to the skill's primary orchestration purpose.
- [Indirect Prompt Injection] (LOW): The skill accepts 'feature_description' and 'tech_stack' as user inputs and interpolates them into agent prompts without delimiters or sanitization. Evidence Chain (SKILL.md): 1. Ingestion points: 'feature_description' and 'tech_stack' fields. 2. Boundary markers: Absent. 3. Capability inventory: Spawning sub-agents (toolbox-resolver, repo-analyst, etc.) and dynamic skill injection. 4. Sanitization: Absent.
Audit Metadata