feature-dev
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied input via the
$ARGUMENTSplaceholder to drive a multi-phase development workflow. - Ingestion points: User input enters via the
$ARGUMENTSvariable inSKILL.mdduring Phase 1 (Discovery). - Boundary markers: The instructions do not define specific delimiters or "ignore instructions" warnings when interpolating the user input into the agent's context.
- Capability inventory: The skill performs codebase analysis using
view_file, modifies files usingwrite_to_file, and orchestrates multiple sub-agents for code exploration, architecture design, and quality review. - Sanitization: There is no explicit sanitization, validation, or escaping of the external input provided in
$ARGUMENTSbefore it is used to influence agent tasks. - Mitigation: The risk is significantly reduced by mandatory human-in-the-loop checkpoints in Phases 3, 4, 5, and 6, where the agent is explicitly instructed to wait for user answers and approval before proceeding with design or implementation.
Audit Metadata