llm-prompt-optimizer
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill provides purely informational guidance and templates for prompt engineering. No executable code, scripts, or automated tool calls are present. The content is limited to markdown documentation for human or AI reference.
- [INDIRECT_PROMPT_INJECTION]: The skill documents methods for interpolating external content into prompts, which is an inherent vector for indirect prompt injection if the resulting prompts are used to process untrusted data.
- Ingestion points: Content placeholders in templates within
SKILL.md(e.g.,[input text here],[your review here],[your context here]). - Boundary markers: The skill correctly identifies this risk and includes a specific recommendation in the 'Best Practices' section to use delimiters (e.g., triple backticks or dashes) to separate instructions from untrusted content.
- Capability inventory: None. The skill contains no executable code, subprocess calls, or tool definitions.
- Sanitization: Not applicable as the skill does not implement processing logic, only documentation.
Audit Metadata