escalation

Decide when and how to escalate a customer conversation to a human agent. Use when a request is high-risk, the customer is frustrated, or the issue is outside what you can resolve.

Quality

87%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

100%Weight 40%Scale 1-3

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a tight, well-organized instruction skill: concise, concrete, and clearly sequenced with no unnecessary explanation. No bundle files exist and none are needed, so progressive disclosure is satisfied by the clean section structure.

Dimension	Reasoning	Score
Conciseness	The body is lean and free of padding; it explains nothing Claude already knows and every line ('Include any order IDs and what you already tried', 'Set expectations... rough timeline') earns its place.	3 / 3
Actionability	Concrete, specific, instruction-only guidance: numbered steps give exact instructions ('Summarize the issue in two or three sentences', 'Include any order IDs and what you already tried'), which is actionable without code.	3 / 3
Workflow Clarity	A clear 1–4 sequence for a simple, non-fragile process; the destructive/batch cap does not apply, and the simple-skills note allows a 3 for an unambiguous single process.	3 / 3
Progressive Disclosure	Under 50 lines with no need for external references, organized into clearly separated 'When to escalate' and 'How to escalate' sections, matching the simple-skill allowance for a 3.	3 / 3
	Total	12 / 12 Passed

Description

75%Weight 40%Scale 1-3

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is concise, uses correct third-person voice, and clearly covers both capability and explicit triggers, with low conflict risk. Its main weakness is generic trigger phrasing and a single named action rather than multiple concrete ones.

Suggestions

Add concrete trigger keywords users would actually say (e.g. 'refund', 'billing dispute', 'speak to a human', 'manager') to strengthen trigger_term_quality.

List a couple of specific concrete actions (e.g. 'Summarize the issue and hand off to a human agent') to lift specificity from a single named action to multiple.

Dimension	Reasoning	Score
Specificity	Names a clear domain and action ('escalate a customer conversation to a human agent') but describes one core action rather than listing multiple concrete sub-actions, so it does not reach the anchor-3 bar.	2 / 3
Completeness	Explicitly answers what it does ('Decide when and how to escalate... to a human agent') and when to use it via an explicit 'Use when...' clause covering high-risk, frustrated, and out-of-scope cases.	3 / 3
Trigger Term Quality	Triggers ('high-risk', 'the customer is frustrated', 'outside what you can resolve') are natural but abstract; they lack the concrete keyword variations a user would actually say (e.g. 'refund', 'billing', 'agent').	2 / 3
Distinctiveness Conflict Risk	The escalation-to-human-agent niche is distinct, and its triggers (high-risk, frustrated, out-of-scope) are unlikely to fire for unrelated skills.	3 / 3
	Total	10 / 12 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: cloudflare/agents
Path: think-starters/customer-support/agents/support/skills/escalation/SKILL.md
Commit: ffc2d95

Reviewed: 12 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.