prompt-optimization

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a prompt repetition technique that creates a surface for indirect prompt injection. By duplicating the user-provided prompt to improve model accuracy, any malicious instructions contained within that prompt are also duplicated, potentially increasing the likelihood of the agent obeying them.
  • Ingestion points: The user-supplied prompt processed by the Haiku agent (SKILL.md).
  • Boundary markers: None; the transformation described uses simple newline separation without delimiters to isolate the original content.
  • Capability inventory: The skill is designed to trigger during high-privilege tasks such as unit test execution and linting (SKILL.md).
  • Sanitization: No sanitization or instruction-filtering mechanism is defined before the prompt is repeated.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 02:28 AM
Security Audit — agent-trust-hub — prompt-optimization