create-meta-prompts
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): The templates and reference files do not contain instructions to bypass safety guidelines or ignore system constraints. The use of XML tags (e.g., , ) provides structural integrity that helps prevent instructional drift.
- [Data Exposure & Exfiltration] (SAFE): There are no patterns suggesting the exfiltration of sensitive data. Path references are localized to the .prompts/ directory or standard project paths (e.g., src/), and no network calls are present in the skill code.
- [Obfuscation] (SAFE): All files are written in clear, human-readable Markdown and XML. No Base64 encoding, zero-width characters, or homoglyph attacks were identified.
- [Remote Code Execution] (SAFE): The skill does not download or execute remote scripts. While the generated prompts may suggest verification commands (e.g., npm test), these are part of a standard development workflow and are not executed automatically by the skill itself.
- [Indirect Prompt Injection] (LOW): The skill provides an interface for processing external data (user descriptions and research files). It mitigates associated risks through mandatory XML boundary markers and structured output schemas.
- Ingestion points: User 'description' input and existing .md files in the .prompts/ folder.
- Boundary markers: Heavy use of XML tags in all provided templates.
- Capability inventory: The skill is limited to file system read/write operations within the project context.
- Sanitization: Relies on structural delimiters to separate data from instructions.
Audit Metadata