yzr-code-refactoring-review
Warn
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions and language-specific plugins direct the agent to execute shell commands that can trigger arbitrary code execution during the build or installation process. For example, 'mvn compile' in 'references/languages/java.md' and 'npm install' in 'references/languages/typescript.md' can execute untrusted code if the target project contains malicious build scripts, plugins, or lifecycle hooks.
- [PROMPT_INJECTION]: The skill facilitates indirect prompt injection by ingesting untrusted source code for analysis without specifying protective boundaries or instructions to ignore embedded commands. Malicious instructions hidden in comments or strings within the reviewed code could influence the agent's behavior or findings.
- Ingestion points: User-provided code segments, file paths, and git diffs (defined in 'SKILL.md' Step 1).
- Boundary markers: None identified; instructions do not specify using delimiters or 'ignore' blocks for the code being reviewed.
- Capability inventory: File system access (reading code), shell command execution (running linters and build tools), and report generation.
- Sanitization: No validation or filtering is performed on the ingested code before processing.
Audit Metadata