manager-setup

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of Bash and Node.js to automate the setup and configuration of the agent's environment. This includes creating project-specific directories, writing state files, and idempotently editing the project's 'settings.local.json' file to register tool-use hooks.
  • [PROMPT_INJECTION]: The skill implements a system of 'codewords' (e.g., ++m, ++a) that, when detected in a user's prompt, trigger the injection of pre-defined behavior blocks. While technically a form of prompt injection, this is the core intended functionality of the skill, used to guide the AI's persona and methodology.
  • [PERSISTENCE_MECHANISMS]: By registering a 'PreToolUse' hook in the local project configuration, the skill establishes a persistent logic gate that monitors and restricts tool execution across agent sessions. This mechanism is transparently documented as the skill's primary method for enforcing delegation.
  • [DYNAMIC_EXECUTION]: The installation and management logic is executed dynamically via Node.js command-line flags ('node -e'). This allows the skill to perform complex JSON manipulation and file synchronization without relying on standard file-editing tools which may be blocked by the skill's own security policies.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 06:23 PM
Security Audit — agent-trust-hub — manager-setup