init-ai-md
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were detected in the skill instructions or templates.
- [COMMAND_EXECUTION]: The skill includes guidelines for generating release logs using the well-known
@changesets/clitool via pnpm dlx. This is a standard operation for project maintenance and version control. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data from documentation files, creating a potential surface for indirect prompt injection. 1. Ingestion points: Project memory files (CLAUDE.md, AGENTS.md, GEMINI.md) and internal markdown templates. 2. Boundary markers (absent): Not explicitly defined in templates; the skill relies on user-driven selection. 3. Capability inventory: File read/write operations and internal command triggers (/init). 4. Sanitization (present): The skill mandates user review and selection via the AskUserQuestion tool before applying any changes, ensuring human-in-the-loop validation.
Audit Metadata