HZL

**Judgment:** HZL is a strong, well-designed task ledger skill that meaningfully upgrades OpenClaw’s ability to handle persistent, multi-step and multi-agent workflows. It is especially valuable wherever work must survive session boundaries or be safely handed off between agents. **Strengths & risks:** - **Strengths:** - Purpose-built for OpenClaw’s ecosystem, with explicit guidance for single- and multi-agent pools. - Rich task model (projects, tasks, subtasks, dependencies, leases, checkpoints) and clear operational patterns. - Good observability and control surfaces (stuck detection, status, events, stats, hooks, dashboard). - **Risks / limitations:** - Contains **destructive CLI flags** (`hzl init --force`, `hzl init --force --yes`, `hzl task prune --yes`) that **must never be used unless the user explicitly requests data deletion**. - Some features (e.g., `hzl workflow run` commands) depend on HZL v2+, so behavior may differ on older installations. - Adds conceptual complexity; improper setup of projects/agents/leases could lead to routing confusion or apparent “lost” tasks. **Recommended scenarios:** - Long-running or non-trivial multi-step work where you expect to pause and resume over multiple sessions. - Multi-agent or multi-role environments where tasks need to be routed via pools and reclaimed if an agent stalls. - Workflows requiring explicit handoff, delegation, and recovery semantics, rather than ephemeral, one-shot commands. Avoid using HZL for trivial one-off actions or time-based reminders; in those cases, the overhead and statefulness are unnecessary and simpler mechanisms (e.g., direct execution or Cron) are preferable.