forge-idiomatic-engineer

Warn

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill directs the agent to locate and execute binary entry points directly from the local repository, such as the 'Forge CLI command', 'project wrapper', or 'checked-in build output'. This pattern allows a malicious repository to supply a compromised binary that the agent would then execute with its active session privileges.
  • [COMMAND_EXECUTION]: The instruction to use the 'lsof' command with a port value parsed from 'forge.toml' (e.g., 'lsof -iTCP:') creates a potential command injection vulnerability if the configuration data contains shell-sensitive characters like semicolons or pipes.
  • [PROMPT_INJECTION]: The skill exhibits a large indirect prompt injection surface. It is designed to ingest and act upon data from external repositories while possessing high-privilege capabilities including code generation, file modification, and command execution.
  • Ingestion points: 'forge.toml', 'Cargo.toml', 'src/main.rs', 'migrations/', and frontend configuration files like 'package.json'.
  • Boundary markers: Absent. No instructions are provided to delineate or ignore potential instructions embedded within the repository data.
  • Capability inventory: The skill utilizes subprocess execution for 'forge', 'cargo', 'bun', 'npm', 'lsof', and 'curl', and performs extensive file system write operations.
  • Sanitization: Absent. The skill encourages direct adoption of patterns and commands found in the repository without validation.
  • [EXTERNAL_DOWNLOADS]: The skill manages project dependencies via standard package managers (npm, bun, cargo) and uses the 'forge test' command, which facilitates the download and installation of Playwright browsers.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 13, 2026, 10:17 PM