meta-prompt-builder
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests natural language task descriptions and orchestration plans to generate instructions for other agents, creating an attack surface for indirect prompt injection.
- Ingestion points: Natural language task inputs and orchestration plans are received in Step 0 (SKILL.md).
- Boundary markers: The skill structures data into specific fields (Goal, Context, Constraints, Done When) but lacks explicit delimiters or instructions to prevent the agent from obeying commands embedded within the input data.
- Capability inventory: The skill is authorized to use Read, Bash, Glob, and Grep tools to verify agent capabilities and harness state (SKILL.md).
- Sanitization: There is no evidence of input sanitization or escaping of the natural language content before it is processed into prompt drafts.
Audit Metadata