impl
Pass
Audited by Gen Agent Trust Hub on Feb 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill dispatches sub-agents to execute tasks, which involves writing code and running tests on the local filesystem. This behavior is the intended primary purpose of the skill for TDD-driven development.
- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it reads task descriptions and state data from local files to populate sub-agent prompts.
- Ingestion points: Reads
{task_id}.mdandstate.jsonfrom the feature output directory. - Boundary markers: None explicitly mentioned in the instructions to the sub-agents to ignore embedded commands.
- Capability inventory: Sub-agents can modify files, execute shell commands (tests), and commit changes via Git.
- Sanitization: No explicit sanitization or validation of the content within task files is described before it is provided to the sub-agent.
Audit Metadata