verify-gate

Warn

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill allows the execution of arbitrary shell and JavaScript code defined in project documentation or configuration via the mcp-scripts feature. This enables project-specific logic to be executed directly by the agent during the verification phase.
  • [COMMAND_EXECUTION]: The agent is instructed to discover and run commands from numerous project files including package.json, Makefile, Cargo.toml, and pyproject.toml. A malicious repository could use these files to trigger unauthorized command execution when the agent attempts to verify the code.
  • [DYNAMIC_EXECUTION]: The skill implements a Fix Loop where the agent generates code patches based on compiler or test errors and then executes verification commands against those generated changes, creating a pattern of runtime script generation and execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes instructions and commands from external, potentially untrusted project files.
  • Ingestion points: Reads CLAUDE.md, AGENTS.md, and various configuration files (e.g., package.json, pyproject.toml).
  • Boundary markers: No explicit boundary markers or 'ignore' instructions are used when parsing these files.
  • Capability inventory: Includes the ability to execute shell commands and JavaScript, and the ability to modify project files.
  • Sanitization: There is no evidence of sanitization or validation of the commands discovered in external files before they are executed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 27, 2026, 09:24 PM