Claude Code agent-in-container orchestration and automation

在自己的 Mac 上搭一套常驻、自愈、走订阅、墙内也能用的自托管 AI 伴侣 · 人看版讲思路，机看版给完整规格 · 文 / 小C & Grace

A zero-latency, local-first runtime firewall for LLMs. Intercept every prompt and response locally to stop data leaks and runaway token costs.

Dynamic Effort Orchestration — a Claude Code + Fable 5 harness that assigns each sub-agent the effort level its task deserves. Zero Opus fallback.

High-fidelity Claude Fable 5 (Mythos) environment emulation and automated multi-agent jailbreak (Pack Hunt) research laboratory.

Codex skill for token-conscious Claude Fable 5 consultations via your own Claude Code CLI

Keep Claude Fable 5 (or Opus) as your daily driver without draining your usage limit — token-frugal multi-agent orchestration plugin for Claude Code: model-tier routing, Requirements Ledger, enforced guard hooks

Proactive 5h usage guard for Claude Code (macOS). Pause before extra wallet on long Fable 5 / subagent runs. /usage-guard skill + external daemon. Not affiliated with Anthropic.

Fan-made FireRed-style Pokemon RPG in vanilla HTML/CSS/JS - built as a single-shot prompt trial of Claude's Fable 5 model. Gen 3 battle mechanics, 20 species, mobile touch controls, zero build step.

Forge Fable 5-native prompts, agents, skills, and harnesses from abstract requests — with fail-closed gates guaranteeing zero Opus fallback.

Point Fable 5 at a codebase — it audits and writes plans any agent can execute. A Fable 5 fork of shadcn/improve.

Enterprise Claude Fable 5 API blueprint & agent scaffolding. Features Claude Fable 5 vs GPT-5 benchmarks, pricing per million tokens, 1M context window optimization, and free trial bypass updates. Built for autonomous Mythos-class models.

Desktop client for Claude Fable 5. Unleash Anthropic's Mythos-class model for complex coding, deep system security audits, and multi-week projects. Free thru June 22.

🌙 The black box recorder for overnight AI agents (built for the Fable 5 era of multi-day autonomous runs) — hash-chained ledgers, checkpoints, and a morning debrief that verifies what your agent claims it did. Logs are claims; replays are proofs.