feature-forge
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]
- The skill processes user-supplied text during the discovery and interview phases to generate specifications. This creates a surface for indirect prompt injection if a user embeds malicious instructions in the requirement descriptions.
- Ingestion points: User responses captured via AskUserQuestions and open-ended interview follow-ups.
- Boundary markers: The prompt logic does not explicitly delimit user input or include instructions to ignore potentially malicious embedded commands.
- Capability inventory: The skill has the capability to write files (specs/*.spec.md) and launch platform subagents using the Task functionality.
- Sanitization: There is no evidence of input validation or content filtering for the user-supplied data.
Audit Metadata