agent-browser
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.
68
84%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Critical
Do not install without reviewing
Quality
Discovery
92%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a strong description that clearly communicates capabilities and provides excellent trigger coverage with natural user phrases. The 'Use when' and 'Triggers include' sections make it highly actionable for skill selection. The only minor weakness is potential overlap with web scraping or testing-specific skills, though the browser automation framing helps mitigate this.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, and automating browser tasks. | 3 / 3 |
Completeness | Clearly answers both 'what' (browser automation CLI for AI agents with specific capabilities listed) and 'when' (explicit 'Use when' clause plus a 'Triggers include' section with concrete user phrases). | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural trigger terms users would say: 'open a website', 'fill out a form', 'click a button', 'take a screenshot', 'scrape data from a page', 'test this web app', 'login to a site', 'automate browser actions'. These are highly natural phrases. | 3 / 3 |
Distinctiveness Conflict Risk | While browser automation is a fairly distinct niche, terms like 'extracting data', 'testing web apps', and 'scrape data' could overlap with web scraping tools, testing frameworks, or data extraction skills. The description could be more explicit about what distinguishes it (e.g., it's a CLI tool, uses a specific browser engine). | 2 / 3 |
Total | 11 / 12 Passed |
Implementation
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a highly actionable and well-structured skill with excellent executable examples and clear workflow patterns. Its main weakness is length — it tries to serve as both a quick-start overview and a comprehensive reference, leading to redundancy (especially around authentication patterns) and a body that could benefit from moving detailed sections into the referenced files. The progressive disclosure structure exists but isn't fully leveraged.
Suggestions
Move the detailed authentication options (5 approaches) and security configuration into references/authentication.md and references/security.md respectively, keeping only the recommended approach inline with a link to alternatives.
Consolidate redundant sections — the 'Handling Authentication' section and the authentication entries under 'Common Patterns' overlap significantly; pick one location and reference the other.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is quite long (~400+ lines) and covers many features comprehensively, but some sections are redundant (e.g., authentication is explained in both 'Handling Authentication' with 5 options AND repeated in 'Common Patterns' with auth vault and state persistence examples). The viewport/device section also repeats information. However, it generally avoids explaining concepts Claude already knows and stays focused on tool-specific syntax. | 2 / 3 |
Actionability | Nearly every section includes fully executable, copy-paste-ready bash commands with concrete examples. Commands use real URLs, specific flags, and show expected output formats (e.g., snapshot output with @e1 refs). The JavaScript eval section even explains shell quoting pitfalls with multiple concrete solutions. | 3 / 3 |
Workflow Clarity | The core workflow is clearly sequenced (Navigate → Snapshot → Interact → Re-snapshot) with explicit validation patterns. The 'Ref Lifecycle' section serves as a critical validation checkpoint, the 'Diffing' section provides verification after actions, and the session cleanup section warns about leaked processes. The 'Timeouts and Slow Pages' section provides explicit wait strategies for error recovery. | 3 / 3 |
Progressive Disclosure | The skill has a well-organized reference table pointing to 7 deep-dive documents and 3 templates, which is good progressive disclosure structure. However, the main SKILL.md itself is very long with substantial inline content (authentication with 5 options, extensive common patterns, security configuration) that could be better split into reference files. The body tries to be both a quick-start guide and a comprehensive reference simultaneously. | 2 / 3 |
Total | 10 / 12 Passed |
Validation
81%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 9 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
skill_md_line_count | SKILL.md is long (646 lines); consider splitting into references/ and linking | Warning |
allowed_tools_field | 'allowed-tools' contains unusual tool name(s) | Warning |
Total | 9 / 11 Passed | |
- Repository
- databricks/devhub
- Commit
0b6297e
- Reviewed
Table of Contents
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.