memory-and-rules
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for processing data from external project files (CLAUDE.md, .claude/rules/, and MEMORY.md), which serves as a potential surface for indirect prompt injection if those files contain malicious instructions.
- Ingestion points: The agent is instructed to load instructions from multiple levels of a project hierarchy, including root files, modular rules in subdirectories, and automated memory files.
- Boundary markers: The reference documentation explicitly notes that file imports (using the @ syntax) are not evaluated when they appear inside code spans or code blocks.
- Capability inventory: The skill enables the agent to read, write, and organize memory files, and to open them in a system editor using the /memory command.
- Sanitization: The skill documents a platform-level security control where the first encounter of an external import in a project triggers a mandatory user approval dialog.
- [SAFE]: The skill functions as an instructional reference for standard features of the Claude Code tool. All file paths, environment variables, and shell commands described (such as /init and /memory) are consistent with the official documentation and the intended use-case of project memory management. The skill identifies itself as a reference and includes links to official documentation for verification.
Audit Metadata