groove-utilities-memory-graduate
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes data from markdown files in the memory directory to update AGENTS.md, which governs future agent behavior. Ingestion points: Reads data from .groove/index.md and files within the /learned/ directory. Boundary markers: Absent in source files, but present in the target AGENTS.md file to delimit the graduated learnings section. Capability inventory: Uses Write and Edit tools to modify session constraints in AGENTS.md. Sanitization: The skill implements manual sanitization by requiring the user to confirm the exact text and trim the input via the AskUserQuestion tool before any filesystem changes occur. This surface is inherent to the skill's primary purpose and is well-mitigated by human-in-the-loop validation.
- [SAFE]: No other malicious patterns, such as data exfiltration, obfuscation, or unauthorized command execution, were identified. The skill's operations are consistent with its documented purpose and are restricted to the local project environment.
Audit Metadata