commit
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes standard Git commands (git diff --cached and git commit) to analyze staged changes and apply the generated message. These are appropriate for the tool's stated purpose.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository through the output of git commands.
- Ingestion points: Output of git diff --cached in the Instructions section.
- Boundary markers: No explicit delimiters or instructions to ignore embedded content are provided.
- Capability inventory: The skill can execute git commit via the shell.
- Sanitization: The skill implements a human-in-the-loop safeguard by requiring user confirmation before executing the commit command.
- [SAFE]: Analysis of the instructions and execution logic revealed no signs of malicious intent, data exfiltration, or obfuscation. The skill's operations are limited to standard local Git workflows.
Audit Metadata