prompt-engineer
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The documentation in
references/system-prompts.mdandreferences/evaluation-frameworks.mdcontains explicit lists of prompt injection payloads, including 'Ignore previous instructions' and 'DAN/jailbreak' patterns. These are used within the skill as part of a diagnostic test suite for evaluating prompt robustness. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of automated evaluation runners that process external data and user-provided prompts, creating an attack surface for indirect injection.
- Ingestion points: User-provided inputs, test cases in JSON format, and external datasets referenced in
references/prompt-optimization.mdandreferences/evaluation-frameworks.md. - Boundary markers: The skill references the use of XML tags (e.g.,
<user_message>) as a defense mechanism but does not enforce them across all code templates. - Capability inventory: The skill provides Python and TypeScript templates for performing network-based LLM API calls and local file system reads.
- Sanitization: Not implemented in the provided code snippets; the skill relies on the end-user to implement proper input validation.
Audit Metadata