What problem does it solve? Setting up an isolated Hermes agent profile per repository involves many fragile steps — profile cloning, skill pruning, project-skill symlinks, hook registration, and project rules — and each step has silent failure modes (corrupted config keys, broken symlinks, ambiguous skill paths) that are hard to diagnose after the fact. ## Core Features & Use Cases - Repo-local profile bootstrap: Creates a ./dev entrypoint with setup, doctor, and passthrough subcommands that never touch the default Hermes profile and stay idempotent across re-runs. - Skill pruning and project-skill symlinks: Clones a profile, disables unwanted skills via a literal JSON list, symlinks project skills into ./.hermes/skills/, and trusts them from the repo root. - Per-profile hooks and project rules: Registers pre_tool_call, post_tool_call, and pre_llm_call hooks, plus AGENTS.md operating rules and a .hermes.md skill routing table with negative controls. - Use Case: You want a frontend repo to run Hermes with only its relevant skills, guarded git operations, and evidence-collecting hooks — run the procedure to mirror an existing profile repo, adapt the disabled-skills list and hooks, then verify with ./dev doctor. ## Quick Start Ask the agent to set up a repo-local Hermes profile with a ./dev entrypoint for this repository, mirroring an existing profile repo and verifying the result with ./dev doctor.
