ralph-execute

Pass

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill directs the agent to perform complex multi-agent orchestration tasks with elevated permissions.\n
  • It explicitly instructs the agent to 'Keep all role configs at full permissions (danger-full-access)'.\n
  • It utilizes agent spawning capabilities with specific isolation settings ('fork_context = true') to manage 'worker' and 'explorer' roles.\n- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection (Category 8) because it uses repository-hosted files to determine orchestration logic.\n
  • Ingestion points: Reads state, policy, and configuration from '.ralph/state/workflow-state.json', '.ralph/state/spec-queue.json', '.ralph/policy/project-policy.md', and 'AGENTS.md'.\n
  • Boundary markers: None identified; the skill instructions mandate treating these files as the 'canonical machine state' and 'source of truth'.\n
  • Capability inventory: Includes the ability to read and write files within the repository and to spawn and configure sub-agents with full access permissions.\n
  • Sanitization: There is no evidence of validation or sanitization for the content of the read files before they are used to influence the agent's workflow and role selection.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 25, 2026, 10:53 PM
Security Audit — agent-trust-hub — ralph-execute