Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
- Setup instructions now clarify that rubric checkpoints should be declarative statements defining expected behavior, replacing yes/no questions. - Added an optional third step: users can configure OpenClaw to send alerts when daily reviews flag issues needing attention. - Alert configuration is now highlighted as entirely optional, with default behavior unchanged (issues surfaced at the next session). - No code or structural changes; documentation improvements only.
What This Skill Does
Wayfound adds structured self-review to an agent's existing memory system. A rubric in SOUL.md defines expected behavior, and a nightly cron job writes a short review to memory/ where it gets picked up naturally. Recurring patterns can update SOUL.md or surface to the user.
Uses the files the agent already reads each session rather than creating a separate evaluation system or directory.
When to use it
- Catching repeated mistakes before they become habits
- Verifying the agent confirmed before making external changes
- Flagging surface-level answers to questions that deserved deeper research
- Surfacing behavioral patterns worth adding as permanent standards
- Reviewing overnight whether any agent actions need user attention
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Set up daily self-review for my agent
- 1Adds a Self-Review Rubric section to SOUL.md with 3-5 specific checkpoints matching the user's context
- 2Checks `openclaw cron list` to confirm no duplicate job exists
- 3Creates a daily cron job at 23:00 running wayfound:daily-review with isolated, cheap model, and low-thinking settings
- 4Explains that issues found are surfaced at the next session start, with optional alert routing available
A rubric is active in SOUL.md and a nightly job writes review files to memory/, compounding findings into future sessions