agent-code-review
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides an interface to process untrusted external content (GitHub Pull Request data) through a code review engine, which presents an indirect prompt injection surface.
- Ingestion points: Pull Request metadata and code contents are retrieved from GitHub and processed by the engine.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified in the skill markdown.
- Capability inventory: The tool can perform automated actions on GitHub (approve, comment, request changes) based on the engine's output.
- Sanitization: No content sanitization or validation logic is described for the PR data before it is sent to the review engine.
- [COMMAND_EXECUTION]: The skill relies on the execution of the
agent-code-reviewbinary and supporting tools likeghandduckdb. Usage is appropriately scoped within theallowed-toolsconfiguration, and command arguments are constrained to the binary's subcommand structure. - [DATA_EXPOSURE]: The skill accesses repository configuration and reviewer allow-lists stored in
~/.config/agent-code-review/config.json. This access is restricted to the data required for the tool to function.
Audit Metadata