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
Risk Level
SAFE
Analyzed
Sep 4, 2026, 08:19 PM
Security Audit — agent-trust-hub — quality-gates