do-git-safe-pr-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests untrusted user data (the 'intent' string) and passes it directly to a command-line interface. 1. Ingestion points: The user-provided 'intent' is passed as a CLI argument in SKILL.md to the 'no-mistakes' tool. 2. Boundary markers: None provided to delimit the input string or warn the agent about embedded instructions. 3. Capability inventory: Shell execution of git, GitButler ('but'), and no-mistakes. 4. Sanitization: No instructions to escape or validate user input before interpolation.
- [COMMAND_EXECUTION]: The skill relies on the presence and execution of external CLI utilities including GitButler and no-mistakes, involving shell command construction with dynamic parameters.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to assemble shell commands at runtime incorporating user-supplied strings, creating a surface for command injection if the agent environment does not properly handle argument escaping.
Audit Metadata