groove-utilities-memory-mistakes

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by recording user-provided mistake descriptions and root causes into the agent's persistent memory files.
  • Ingestion points: User-provided descriptions and root causes captured through the $ARGUMENTS variable or interactive prompts in SKILL.md.
  • Boundary markers: Absent; the skill appends user content to markdown files in the .groove/memory/learned/ directory without using delimiters to separate untrusted data from the agent's instructions.
  • Capability inventory: The skill has access to file system operations (Read, Write, Edit) and shell execution (restricted Bash access to git and beans) as specified in the SKILL.md file.
  • Sanitization: No sanitization, validation, or escaping of the user-provided text is performed before it is written to the persistent storage.
  • [COMMAND_EXECUTION]: The skill invokes external command-line utilities to manage tasks and version control.
  • Evidence: Uses the Bash tool with restricted prefixes to execute git commands and beans commands for task tracking, including listing, creating, and updating incidents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 09:04 AM