grace-init
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill populates project documentation and context files (AGENTS.md, XML files) using user-supplied strings. These files are later intended to be read by AI agents to define their operational rules and technical context. While standard for a bootstrapper, this allows for indirect prompt injection if the input contains malicious instructions designed to be followed by an agent in future sessions.
- Ingestion points: Project name, annotation, keywords, and goals provided by the user during the bootstrap process.
- Boundary markers: None used within the XML or Markdown templates to delimit user input from structural instructions.
- Capability inventory: Local file system writes to create the .grace directory structure and project files.
- Sanitization: Placeholders in the templates are replaced directly with user input without validation or escaping.
Audit Metadata