Agent Escalation Manager

You were halfway through a workflow when it started breaking. A login timed out, the next attempt failed again, the agent switched tools with no plan, asked you a vague question, and kept pushing after the run should have paused. Now the job is slower, riskier, and harder to recover than the original failure ever was.

Agent Escalation Manager gives your OpenClaw agent a practical failure-decision system. It classifies what actually went wrong, chooses whether to retry, switch tools, hand off, pause, or stop, explains why that level applies, and returns a clean next-best action ladder with the exact handoff, pause, or stop note needed to recover without more chaos.

What's Included

• complete self-contained SKILL.md
• 6-level escalation ladder
• failure classification model
• install verification workflow
• human handoff template
• pause template
• stop template
• common failure-pattern library

Why This is Better

• Most tools focus on one narrow step — this covers the full post-failure decision ladder
• Typical troubleshooting prompts only suggest retries — this forces retry vs switch vs handoff vs pause vs stop
• Most workflow guides are theory-heavy — this is built for real OpenClaw runs under pressure
• Typical handoff flows produce vague asks — this gives one exact human step and what happens after
• Most recovery tools treat all failures the same — this classifies the failure before choosing the path
• Built to work on first install with zero configuration

Works Well With

• Human Approval Ladder
• Browser Automation Rescue Kit
• Skill Stack Router