gate-wf
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell commands (git, gh, shasum, sed, awk, jq) to compute repository identifiers, generate scoped diffs, and manage its state and cache. These commands are executed within a restricted temporary directory and are used for their intended analytical purposes.
- [EXTERNAL_DOWNLOADS]: Instructions are provided for the user to manually install optional reviewer skills from external GitHub repositories. The skill does not perform unauthorized background downloads or executions from untrusted sources.
- [DATA_EXFILTRATION]: The skill collects project-specific data, such as code diffs and pull request information from the GitHub and Linear APIs. This context is ingested solely to provide information to the review agents and is not transmitted to unauthorized third-party services.
- [PROMPT_INJECTION]: The skill processes untrusted data (code changes, documentation, and external comments) which presents an indirect prompt injection surface. This is mitigated by orchestrating the agents with fixed system prompts, strict JSON output schemas, and instructions to remain in a read-only mode.
Audit Metadata