tdd-cycle
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted requirements from project files, creating a surface for indirect prompt injection.
- Ingestion points: Task descriptions are read from
project/TASKS.mdand user-provided requirements in Phase 1. - Boundary markers: The instructions do not define delimiters or protective instructions to separate ingested data from agent logic.
- Capability inventory: The skill can modify the file system (Steps 2, 4, 6, 8), execute shell commands for testing (Steps 3, 5, 7), and call the
self-heal.skill. - Sanitization: No validation or sanitization is performed on the requirements before they are used to generate code.
- [COMMAND_EXECUTION]: The skill executes shell commands obtained from project-level configuration files such as
AGENTS.md. While functional for TDD, this provides a mechanism for executing commands defined within the local workspace environment.
Audit Metadata