prompt-engineer

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The reference materials in references/system-prompts.md and references/evaluation-frameworks.md contain explicit strings used for prompt injection attacks, including "Ignore all previous instructions" and "reveal your system prompt." These are intended for educational and defensive testing purposes.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to interpolate external user data into instructions and evaluation scripts.
  • Ingestion points: External data enters the agent context through template variables like {{review}}, {{document}}, and {input}.
  • Boundary markers: The skill documentation suggests using XML-style tags and instruction hierarchy as defensive delimiters.
  • Capability inventory: The reference files provide Python and TypeScript code examples for automated LLM calls, tool-using ReAct loops, and automated score aggregation.
  • Sanitization: Defensive strategies such as input sandboxing, canary tokens, and response validation are documented in the reference guides.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 09:07 PM