agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, verify dev server output, test web apps, navigate pages, fill forms, click buttons, take screenshots, extract data, or automate any browser task. Also triggers when a dev server starts so you can verify it visually.

Quality

85%

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

82%Weight 40%Scale 1-5

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

A highly actionable, well-structured CLI reference: every section is executable and the core workflow plus ref-lifecycle rule give clear sequencing. Weaknesses are mild — repeated command examples across sections, no explicit error-recovery guidance, and a large inline command reference that could be split into a bundle file.

Suggestions

Consolidate the repeated open/wait --load networkidle/snapshot -i examples that appear in Dev Server Verification, Command Chaining, and Timeouts into the Core Workflow to trim token usage.

Add a short error-recovery note (e.g., what to do when a ref is stale, an element isn't found, or a wait times out) to close the validation gap in the workflow.

Consider moving the full Essential Commands list and rarer patterns (PDF capture, diff, recording) into a references/ file, keeping SKILL.md as the workflow-focused overview.

Dimension	Reasoning	Score
Conciseness	The body is a lean, command-first reference with essentially no explanation of concepts Claude already knows. It is not a 5 because command examples (open/wait --load networkidle/snapshot -i) repeat across the Dev Server Verification, Command Chaining, and Timeouts sections and could be consolidated.	4 / 5
Actionability	Every snippet is an executable, copy-paste-ready command, and the pattern sections (form submission, auth with state persistence, data extraction, visual debugging) cover the common cases. Nothing is pseudocode or abstract.	5 / 5
Workflow Clarity	The four-step core workflow (Navigate → Snapshot → Interact → Re-snapshot) is clearly sequenced, and the Ref Lifecycle section with 'MUST re-snapshot' and the wait guidance act as checkpoints. It is not a 5 because there is no explicit error-recovery guidance for failure modes like stale refs, missing elements, or timeouts.	4 / 5
Progressive Disclosure	The single file is well-sectioned with clear headers and no nested references, so navigation is easy. It is not a 5 because ~100 lines of the body are an inline command/pattern reference with no bundle files or split-out advanced material, which is at the edge of what belongs in SKILL.md itself.	4 / 5
	Total	17 / 20 Passed

Description

88%Weight 40%Scale 1-5

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

A strong description: third-person, concrete action list, explicit 'Use when' triggers, and a distinctive named-tool niche. The only gap is a few missing natural synonyms (scraping, login flows, e2e) that would push trigger coverage to comprehensive.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete actions — "navigate pages, fill forms, click buttons, take screenshots, extract data, verify dev server output, test web apps" — giving comprehensive coverage of the browser-automation domain. It is not a 4 because coverage goes beyond 'minor gaps' to enumerate essentially all common browser tasks.	5 / 5
Completeness	Both parts are explicit: "Browser automation CLI for AI agents" plus the action list answers 'what', and "Use when the user needs to..." with concrete trigger phrases plus the additional dev-server trigger answers 'when'. Not a 4 because the 'when' is fully explicit and multi-triggered, not merely adequate.	5 / 5
Trigger Term Quality	Good natural-phrase coverage: "interact with websites", "test web apps", "fill forms", "take screenshots", "dev server" are phrases users would actually say. It is not a 5 because common synonyms/variants like "scrape", "log in", or "e2e/end-to-end" are absent.	4 / 5
Distinctiveness Conflict Risk	The browser-automation niche is clear and the tool is named, minimizing wrong-skill triggering. Minor overlap risk remains with web-fetch/scraping and dev-server verification skills via phrases like "extract data" and "verify dev server output", so it is not a 5.	4 / 5
	Total	18 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning

	Total	14 / 16 Passed

Repository: openai/plugins
Path: plugins/vercel/skills/agent-browser/SKILL.md
Commit: 5fd93af

Reviewed: 2 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.