workflow-feature
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided feature descriptions to generate technical implementation checklists (WORKFLOW_*.md), which creates a surface for indirect prompt injection if the user input contains malicious instructions.\n
- Ingestion points: Untrusted user input (feature requests) and local project configuration files including .agents/workflow.json and .agents/AGENTS.md.\n
- Boundary markers: The skill instructions do not specify the use of delimiters or specific boundary warnings when the agent interpolates user input into the generation of the workflow checklist.\n
- Capability inventory: The skill possesses the capability to write files to the project root, execute the gradle_sync tool, and invoke other project skills such as compiler and git-governance.\n
- Sanitization: There is no evidence of sanitization, filtering, or validation of the user's feature description before it is utilized to create technical tasks.
Audit Metadata