prompt-architect
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted user input and external task definitions to generate prompts that are subsequently executed by the agent, creating an injection surface.
- Ingestion points: Workflow Step 1 in
SKILL.mdprocesses user-supplied text descriptions and JSON data from previous task clarification steps. - Boundary markers: The skill uses XML tags to structure the generated output, but it lacks specific delimiters or 'ignore embedded instructions' warnings around the interpolated user-provided task descriptions.
- Capability inventory: The generated prompt can be directly executed (Workflow Step 5), which may lead to file modifications or command executions based on the instructions inside the prompt.
- Sanitization: There is no evidence of sanitization or validation of the input strings before they are interpolated into the Role, Goal, and Task sections of the prompt template.
Audit Metadata