Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
- Added _meta.json file to the skill package. - No changes to skill behavior or documentation content.
What This Skill Does
A runbook for using OpenClaw's exec tool safely in real workspaces. Covers sandbox vs host execution, working directory discipline, long-running process patterns, and macOS permission issues. Includes concrete playbooks for common failure scenarios like skills not loading or Peekaboo bridge problems.
Instead of generic terminal advice, it gives workspace-specific patterns with explicit safety gates, so destructive commands require confirmation and sandbox/host environment differences don't cause silent failures.
When to use it
- Debugging why a custom skill isn't loading in OpenClaw
- Running untrusted build scripts in a sandbox environment
- Diagnosing macOS screen recording permission failures in OpenClaw
- Polling long-running test or build processes without blocking
- Auditing a workspace before making destructive file changes
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: why isn't my new skill showing up in OpenClaw?
- 1Run `ls -la ./skills` to verify the skill folder exists in the workspace
- 2Run `find ./skills -maxdepth 2 -name SKILL.md -print` to confirm SKILL.md is present
- 3Check frontmatter in SKILL.md for validity
- 4Confirm `skills.load.watch` is enabled if files were recently changed
- 5Report which precedence layer (workspace vs bundled) is winning
Identified missing SKILL.md or invalid frontmatter; user corrects the file and skill loads on next watch cycle.