commit-work

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the git CLI suite (including status, diff, add, restore, and commit) to manage version control. It also directs the agent to execute project-specific verification commands, such as unit tests, linters, or build scripts, as a standard part of the pre-commit workflow.
  • [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading and analyzing untrusted content from the project's working tree and diffs to generate commit descriptions. This creates a potential surface for injection if project files contain instructions designed to manipulate the agent.
  • Ingestion points: The agent ingests external data through git diff and git status (SKILL.md).
  • Boundary markers: The skill does not specify delimiters to isolate ingested file content from agent instructions.
  • Capability inventory: The agent possesses the capability to execute git commands and arbitrary project-defined scripts for verification.
  • Sanitization: The skill does not implement specific sanitization for content read from the repository's diffs before processing it for commit message generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 08:52 PM
Security Audit — agent-trust-hub — commit-work