Add OpenClaw ecosystem: self-healing, memorybox, self-evolving

[Ramsbaby]

Hi! Requesting addition of three production-tested Claude Code tools from the OpenClaw ecosystem:

---

1. openclaw-self-healing (★32) — Tools & Utilities

Link: https://github.com/Ramsbaby/openclaw-self-healing

4-tier autonomous crash recovery for any Claude Code service. Preflight config validation → instant KeepAlive restart → HTTP watchdog with exponential backoff → AI emergency recovery (LLM-agnostic: Claude/GPT-4/Gemini/Ollama) → multi-channel alerts. 64% of incidents resolved autonomously in production. Includes Prometheus metrics, Docker Compose, one-line installer. MIT.

---

2. openclaw-memorybox (★8) — Tools & Utilities

Link: https://github.com/Ramsbaby/openclaw-memorybox

Memory hygiene CLI for Claude Code. Prevents context overflow crashes by keeping MEMORY.md / CLAUDE.md lean. memorybox doctor ~/.claude diagnoses; memorybox split fixes interactively. -83% memory size, -91% context pressure, 0 crashes in production. Zero dependencies, works offline. MIT.

---

3. openclaw-self-evolving (★2) — Agent Skills

Link: https://github.com/Ramsbaby/openclaw-self-evolving

Weekly self-improvement pipeline. Scans ~/.claude/logs/ JSONL files for retry loops, rule violations, user frustration — generates exact CLAUDE.md before/after diffs. Zero API calls, pure bash + python3. Human approval required; false positive rate ~8%. MIT.

---

All three are MIT licensed and battle-tested on a 24/7 production instance. They solve complementary problems: crash recovery, memory hygiene, and behavioral learning — the three pillars of a stable long-running agent.