parallel-code-review
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill establishes a secure multi-agent workflow by explicitly requiring all review subagents to use the
readonly: trueparameter and theexploretool type. This design principle adheres to the least-privilege model, ensuring that the AI agents can analyze repository contents without the ability to execute code, modify files, or perform network operations.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, potentially untrusted source code and git diffs, creating an indirect prompt injection surface.\n - Ingestion points: The skill ingests repository file contents and git diff output through the parallel subagents described in
SKILL.md.\n - Boundary markers: The provided prompt templates for the subagents do not define explicit isolation boundaries or delimiters for the code under review.\n
- Capability inventory: All subagents are strictly limited to read-only file system access (
readonly: true) and analysis tools (subagent_type: "explore"). The main agent uses standardgitCLI commands for scope identification. No execution or modification capabilities are assigned.\n - Sanitization: The workflow does not incorporate specific sanitization or filtering of the source code content before it is processed by the subagents.
Audit Metadata