skills/pitzcarraldo/skills/commit/Gen Agent Trust Hub

commit

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • COMMAND_EXECUTION (SAFE): The skill runs Git commands to manage version control. It uses quoted HEREDOCs ('EOF') to safely handle multi-line commit messages, which prevents the shell from interpreting special characters or performing command substitution within the message body.\n- PROMPT_INJECTION (LOW): The skill is vulnerable to indirect prompt injection because it analyzes untrusted content from the repository (diffs and logs) to generate messages.\n
  • Ingestion points: git diff --cached and git log --oneline -5 ingest code and history from the local repository.\n
  • Boundary markers: Absent. The skill does not delimit the diff content or instruct the agent to ignore instructions within the diff.\n
  • Capability inventory: git commit is used to modify the local repository state.\n
  • Sanitization: Shell-level protection is implemented via quoted HEREDOCs, but no semantic sanitization of the input text is present to prevent instruction injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:13 PM