commit-work

Pass

Audited by Gen Agent Trust Hub on May 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell-based Git commands including git status, git diff, git add, git commit, and git push to manage repository states. These commands are necessary for the skill's primary function of automating version control tasks.\n- [PROMPT_INJECTION]: The skill processes untrusted data from the working directory through git diff output, which is then used by the agent to determine commit logic and messages. This creates a surface for indirect prompt injection where malicious code comments or content could influence the agent's behavior during the commit process.\n
  • Ingestion points: git diff and git diff --cached (SKILL.md)\n
  • Boundary markers: Absent; diff content is processed without explicit delimiters or instructions to ignore embedded commands.\n
  • Capability inventory: Full access to git commit and git push through the spawned agent brief.\n
  • Sanitization: Absent; the agent relies on its internal reasoning rather than programmatic filtering of the diff content.\n- [DATA_EXFILTRATION]: The skill performs git push operations to move code to remote servers. While this is a standard developer workflow, it constitutes data transmission. The skill includes a specific guardrail instruction to detect and stop the process if secrets or tokens are identified in the changes.
Audit Metadata
Risk Level
SAFE
Analyzed
May 12, 2026, 10:36 AM
Security Audit — agent-trust-hub — commit-work