tdd-workflows-tdd-refactor
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by interpolating untrusted data into a high-capability agent context.
- Ingestion points: The prompt template in
SKILL.mddirectly includes$ARGUMENTS, which represents the user-provided code to be refactored. - Boundary markers: The instructions lack explicit delimiters (like triple backticks or XML tags) or system-level warnings to the agent to disregard any natural language instructions found within the
$ARGUMENTSblock. - Capability inventory: The
tdd-orchestratorsubagent is tasked with refactoring code and running tests, which necessitates capabilities such as file modification and shell command execution. - Sanitization: There is no evidence of input validation or sanitization to filter out malicious instructions embedded in the code provided for refactoring.
Audit Metadata