carve-it
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it requires the agent to read and interpret commit diffs and file contents during its analysis and audit phases. Malicious instructions embedded in the code could potentially influence the agent's behavior.\n
- Ingestion points: Commit diffs and file contents are read in Phase 1 (Analysis) and Phase 3 (Purity Audit) of SKILL.md.\n
- Boundary markers: The instructions do not define specific delimiters or instructions to distinguish between code-as-data and code-as-potential-instructions during the audit.\n
- Capability inventory: The skill possesses the ability to execute Git commands (checkout, add, commit, branch, switch), write to the filesystem, and run project-specific linting and testing tools.\n
- Sanitization: There is no mention of sanitizing or escaping the content of the files being reviewed before the agent processes them.\n- [COMMAND_EXECUTION]: The skill performs command execution based on the local environment. In Phase 3 (Unit verification), the agent is instructed to infer and run the project's lint and test commands. This creates a risk if the repository contains malicious scripts or configurations that are triggered by these standard development tools.
Audit Metadata