git-master

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes numerous local shell commands to manage Git repositories, including operations that modify history such as git rebase, git reset --hard, and git push --force-with-lease. These operations are fundamental to the skill's purpose as a Git automation assistant and are accompanied by logic-based safety guidelines for the agent.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and analyzes potentially untrusted data from the Git history to determine commit styles and languages.\n
  • Ingestion points: In SKILL.md, the agent executes git log -30 and git blame to extract commit messages and author information for analysis.\n
  • Boundary markers: The skill does not implement boundary markers or instructions to ignore potential commands embedded within the commit messages it reads.\n
  • Capability inventory: The skill possesses significant capabilities, including the ability to execute shell commands, rewrite history, and push changes to remote servers.\n
  • Sanitization: There is no evidence of sanitization or filtering applied to the commit messages before the agent processes them.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 03:50 PM