quick-design
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses internal workflow markers such as 'Verdict: REDIRECTED' and 'Verdict: COMPLETE'. These are used for logical branching within the instructions and do not attempt to override the agent's safety constraints.
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection. 1. Ingestion points: The skill reads from user-provided 'argument-hint' and existing files in 'design/gdd/'. 2. Boundary markers: The skill does not use specific delimiters or 'ignore embedded instructions' warnings when interpolating external content into its design templates. 3. Capability inventory: The skill utilizes 'Write' and 'Edit' tools to modify the local project directory. 4. Sanitization: There is no evidence of filtering, escaping, or validation logic for data read from external sources. This surface is inherent to the skill's primary function as a documentation generator.
Audit Metadata