beastmode-pilisted
Install: claude install-skill lac5q/beastmode
# Beastmode Pi — Pi Harness Adapter
This skill is **only the harness mechanics**. The framework, tier-routing
rule, verifier-first design principle, and self-improvement loop live in the
canonical `beastmode` skill (v2.1) — load it first and follow it. This
adapter tells you which pi packages implement which role and how to wire
them together.
| Universal Beastmode role | Pi implementation |
|---|---|
| Director / Lead (frontier tier) | The pi session itself — frontier model stays in charge |
| Watcher (adversarial review) | `pi-dynamic-workflows` `verify()` / `judgePanel()`, or the built-in `/adversarial-review` workflow |
| Worker (economy tier) | `pi-dynamic-workflows` `agent()` with `tier: "small"`/`"medium"` or an external lane (Qwen / MiniMax API / Droid MiniMax) routed via `model:` |
| Loop engine (continues until done) | `@narumitw/pi-goal` — `/goal`, `goal_complete`, `goal_blocked` |
| Anti-spin circuit breaker | `pi-loop-police` (passive, always on) |
| Worker-contract enforcer | `@gotgenes/pi-permission-system` project config (path / external_directory / bash deny + ask surfaces) |
| Live progress visibility | `@juicesharp/rpiv-todo` overlay (TUI) and `/workflows` navigator |
| Remote supervision | `@llblab/pi-telegram` proactive push (only if `/telegram-setup` was completed) |
| Durable goal record | Optional: any goal store (MemRoOS `/api/gsd/goal`, plain file, etc.). Beastmode itself does **not** require a specific store. |
| Self-improvement log | `<repo>/.le