Define what gets escalated, to whom, and how fast, so problems reach a decision maker at the right moment rather than too early or too late. Use when issues either sit unresolved or interrupt the wrong people constantly.
Scanned 9/5/2026
Install to Claude Code
npx -y skills add Amey-Thakur/AI-SKILLS --skill agent-escalation-ladder --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Agent Escalation Ladder?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/amey-thakur-agent-escalation-ladder)More formats (shields.io, HTML) on the badges page.
---
name: agent-escalation-ladder
description: Define what gets escalated, to whom, and how fast, so problems reach a decision maker at the right moment rather than too early or too late. Use when issues either sit unresolved or interrupt the wrong people constantly.
---
# Agent escalation ladder
Without a ladder, escalation is set by temperament: the anxious escalate
everything and the confident escalate nothing until it is a crisis. A
written ladder replaces both with a rule, which is the only thing that
works when many agents run unattended.
## Method
1. **Define the triggers, not the feelings.** Time elapsed, threshold
crossed, decision required, or risk above a level. Each trigger names
the next rung explicitly.
2. **Escalate with a recommendation, never a raw problem.** What
happened, what was tried, the options, and what you would do. An
escalation that only reports is delegation of thinking upward.
3. **Set the speed by consequence.** A customer-affecting failure moves
immediately; a stalled internal task waits for the weekly cycle.
Treating both as urgent destroys the signal.
4. **Escalate one rung at a time.** Skipping levels except in genuine
emergency prevents the person above from being the queue for
everything.
5. **Make anything binding stop at a human.** Money, contracts,
customer communication, and legal exposure are always the top rung,
whatever the size (see agent-company-blueprint).
6. **Record what was escalated and how it resolved.** Repeated
escalations of the same kind indicate a missing rule or a missing
capability rather than bad luck.
7. **De-escalate explicitly too.** An issue that turned out minor should
be closed loudly enough that everyone stops watching it.
## Boundaries
The ladder routes decisions; it does not make them, and the top of every
ladder is a human. It cannot judge severity beyond the criteria written
into it, so novel situations will be misrouted and need human review.
Incidents involving safety, legal exposure, or personal data have
mandatory paths that override any internal ladder (see
agent-crisis-comms).
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!