发现优质的 AI Agent 技能

来自 openclaw/skills 技能

Sandboxed installation pipeline for ClawHub skills. Install → Stage → Scan → Promote or Quarantine.

A battle-hardened watchdog that keeps your OpenClaw gateway running — and tells you when it can't.

> When retrieving one topic, follow cross-references to load related context. > This mimics spreading activation: activating one concept primes associated concepts. > Updated by nightly consolidation.

> What I know, how well I know it, and where the gaps are. > Confidence: 🟢 High (0.8+) | 🟡 Medium (0.5-0.8) | 🔴 Low (<0.5) > Updated by nightly consolidation. Reviewed weekly.

> Commitments, scheduled actions, and deferred intentions. > Scanned every session start + every heartbeat. Execute when triggered. Archive when done. > > Format: > ```

Schedule: `0 4 * * 0` (Sunday 4 AM) Recommended model: Any model with reliable tool use Timeout: 600 seconds

Copy-paste these into your AGENTS.md.

Schedule: `0 2 * * *` (2 AM daily) Recommended model: Any model with reliable tool use (Sonnet 4+, GPT-5+) Timeout: 300 seconds

Add these settings to your `openclaw.json` under `agents.defaults`.

Give your agent a mind, not just memory.

> Insights, corrections, and best practices. Promote recurring patterns to SOUL.md or AGENTS.md.