structured-autonomy-plan
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves researching user-requested features, external APIs, and code patterns via the
#tool:runSubagenttool or other search tools. This creates a surface for indirect prompt injection if the researched content (e.g., third-party documentation or comments in external repositories) contains adversarial instructions intended to influence the agent's planning output. - Ingestion points: User feature requests and external research findings gathered in Step 1 of the workflow.
- Boundary markers: Absent. The instructions do not specify delimiters or warnings to ignore instructions embedded within researched data.
- Capability inventory: Invocation of subagents and writing files to the local file system.
- Sanitization: Absent. The agent is not instructed to filter or escape the content it retrieves during research.
- [COMMAND_EXECUTION]: The workflow requires the agent to save generated plans to the local filesystem at
plans/{feature-name}/plan.md. This represents a file-write capability that relies on user-influenced input for the directory and filename structure.
Audit Metadata