write-tests

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests content from local files and git diffs and interpolates them directly into templates for sub-agents, creating an indirect prompt injection surface.
  • Ingestion points: Local altered files ({FILE_PATH}) and git diff data ({GIT_DIFF_OUTPUT}) are read during the analysis and test creation steps in SKILL.md.
  • Boundary markers: Absent; the agent templates do not include robust delimiters or specific instructions to ignore malicious prompts hidden within code comments or string literals.
  • Capability inventory: Spawns subordinate agents (Haiku, Sonnet, Opus models) and executes test suites locally via a dynamically identified test command ({TEST_COMMAND}).
  • Sanitization: Absent; no escaping or sanitization is performed on the repository contents before they are interpolated into instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 01:10 AM
Security Audit — agent-trust-hub — write-tests