What problem does it solve? Delegating narrow subtasks to a local AI specialist risks accidental commits, secret leakage, or uncontrolled system changes. This Skill enforces a bounded execution contract so Hermes handles PT-orama subtasks locally without committing, deploying, or exposing private state. ## Core Features & Use Cases - Bounded Execution Contract: Restates the subtask, identifies canonical source docs, and returns focused output in a fixed response shape (SUBTASK, ASSUMPTIONS, RESULT, VERIFICATION, RISKS, NEXT HANDOFF). - Safety Boundaries: Prohibits commits, deletions, deploys, force-pushes, provider changes, and any copying of private Hermes home state or secrets into repositories. - Tool Dispatch Rules: Verifies LM Studio with a chat-completions canary, installs AGY via the official installer when absent, and routes mechanical Codex work through the codex-cli dispatch reference without hardcoded host paths. - Use Case: When Hermes is asked to run a small PT-orama subtask locally on Windows with LM Studio, this Skill ensures the work stays scoped, verified, and free of private data leakage. ## Quick Start Ask Hermes to handle a narrow PT-orama subtask as a local specialist using the pt-orama-delegate command.
