quality-gates
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill outlines a workflow that ingests potentially untrusted data from user feedback and external test logs, creating an attack surface for indirect prompt injection.
- Ingestion points: Data enters the context via user-provided validation feedback (Pattern 5, Step 4) and captured stdout/stderr logs from test executions (Pattern 6, Step 2).
- Boundary markers: The skill lacks instructions for using delimiters or boundary markers when interpolating this external data into agent prompts, using raw placeholders like
[user's exact words]and[test failure logs]instead. - Capability inventory: The described workflows involve shell command execution (
bun test) and the ability to modify project source code based on agent analysis. - Sanitization: The skill recommends an intermediate step of parsing user feedback into 'structured issues' (Step 5), which provides a degree of validation, though no specific escaping or sanitization techniques are detailed.
- [COMMAND_EXECUTION]: The Test-Driven Development (TDD) loop pattern (Pattern 6) explicitly instructs the agent to execute shell commands such as
bun test tests/auth.test.ts. While standard for local development tasks, this is the primary mechanism through which the ingested data described above can influence system behavior.
Audit Metadata