commit-message

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a Python script scripts/analyze_git.py which executes shell-based Git commands using subprocess.run(shell=True). While the commands are hardcoded in the script, this capability is used to extract repository information.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of repository state.\n
  • Ingestion points: Git diff statistics and file names are read from the local repository state in scripts/analyze_git.py.\n
  • Boundary markers: The skill lacks explicit delimiters or instructions to ignore potential commands embedded within file names or Git diff data.\n
  • Capability inventory: The skill is capable of executing Git commands via the provided script and suggesting further commands to the user.\n
  • Sanitization: There is no evidence of sanitization or escaping of the Git metadata before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 12:52 PM