requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local
gitcommands, includinggit rev-parse,git log, andgit diff, to identify commit ranges and extract source code changes for analysis. These operations are performed within the local environment where the repository resides.- [COMMAND_EXECUTION]: A potential risk of shell command injection exists incode-reviewer.md. The template interpolates{BASE_SHA}and{HEAD_SHA}variables directly into shell commands:git diff --stat {BASE_SHA}..{HEAD_SHA}. If these variables are populated with malicious shell metacharacters (e.g.,;,&&, or|) from an untrusted source, it could lead to unauthorized command execution.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because the code reviewer subagent processes untrusted data (source code changes) which may contain malicious instructions hidden in comments or strings. - Ingestion points: Source code content retrieved via
git diffcommands incode-reviewer.md. - Boundary markers: The template uses Markdown headers to separate sections but lacks specific delimiters or "ignore instructions" warnings to wrap the untrusted code content.
- Capability inventory: The subagent has the capability to execute shell commands and generate structured review feedback.
- Sanitization: There is no evidence of sanitization or escaping of the diff output or the git identifiers before processing.
Audit Metadata