commit

Pass

Audited by Gen Agent Trust Hub on Jul 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes several git commands including git status, git diff, git add, and git commit. These operations are standard for the skill's purpose and are executed within the local repository context.
  • [PROMPT_INJECTION]: The skill reads repository data (git diffs and status) which serves as a surface for indirect prompt injection if an attacker places malicious instructions in code comments. This is mitigated by a human-in-the-loop requirement for all state-changing actions.
  • Ingestion points: Results from git diff and git status commands defined in SKILL.md.
  • Boundary markers: No specific delimiters are specified for separating diff data from instructions.
  • Capability inventory: The skill can stage files (git add), apply cached patches (git apply), and create commits (git commit).
  • Sanitization: The workflow includes an explicit step to flag secrets, .env files, and credentials to prevent accidental staging.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 7, 2026, 12:13 AM
Security Audit — agent-trust-hub — commit