tdd-loop
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external requirements (tickets, issue descriptions) and has the capability to write code and execute shell commands (test runners). While no active injection is present, the ingestion of untrusted data constitutes an attack surface.
- Ingestion points: User-provided tickets and issue descriptions (SKILL.md).
- Boundary markers: Instructions require restating requirements as observable behavior and include an "Autonomy boundary" that halts execution for changes touching authentication, payments, or data migrations (SKILL.md).
- Capability inventory: Modifying files during the "Green" phase and executing arbitrary test commands defined in project manifests (SKILL.md, references/test-commands.md).
- Sanitization: Relies on structured summary evidence and human-in-the-loop verification gates.
- [COMMAND_EXECUTION]: The skill instructions and reference materials describe the execution of shell commands for running unit and integration tests across multiple programming environments (references/test-commands.md). This is an essential component of the TDD lifecycle.
Audit Metadata