write-commit-message
Pass
Audited by Gen Agent Trust Hub on Feb 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute standard git commands including
git status,git diff, andgit commit. The provided examples correctly use bash heredocs to prevent common command injection vulnerabilities when handling multi-line commit message strings. - [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface as it processes untrusted data from
git diffoutput to generate commit messages. 1. Ingestion points: The output ofgit diffis read and analyzed by the agent. 2. Boundary markers: None are specified to delimit code changes from instructions. 3. Capability inventory: The agent has access to thegitcommand suite. 4. Sanitization: No sanitization is performed on the diff content. As this ingestion is core to the skill's primary function, the risk is categorized as safe. - [SAFE]: No evidence of malicious code, data exfiltration, obfuscation, or persistence mechanisms was found within the skill's instructions.
Audit Metadata