do-commit-push
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple external CLI tools including
git,but(GitButler), andno-mistakes. These tools are used for repository inspection, committing changes, and triggering deployment/publication workflows viaaxi. - [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface.
- Ingestion points: The skill processes untrusted data from repository diffs (
git diff,but diff) and user-provided "intent" descriptions that describe the purpose of the commit. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the content of the diffs or the intent strings as data rather than instructions.
- Capability inventory: The skill has the capability to execute shell commands across several different tools (
git,but,no-mistakes) based on the processed data. - Sanitization: The skill lacks sanitization or escaping mechanisms when interpolating the user-provided intent string into the shell command
no-mistakes axi run --intent "...", creating a potential path for command injection or instruction override.
- Ingestion points: The skill processes untrusted data from repository diffs (
Audit Metadata