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. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.
71
71%
Does it follow best practices?
Impact
55%
4.58xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
Optimize this skill with Tessl
npx tessl skill review --optimize ./.agents/skills/agent-browser/SKILL.mdQuality
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, well-structured description that excels at specificity and trigger term coverage, with an explicit 'Use when' clause and quoted user phrases that make selection easy. Its main weakness is scope creep—by claiming Electron apps, Slack messaging, cloud browsers, and QA testing, it risks conflicting with more specialized skills in those domains. The description uses proper third-person voice throughout.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists many specific concrete actions: navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, automating browser tasks, checking Slack unreads, sending Slack messages, and more. | 3 / 3 |
Completeness | Clearly answers both 'what' (browser automation CLI for AI agents with specific capabilities listed) and 'when' (explicit 'Use when' clause with detailed trigger scenarios and a 'Triggers include' section with quoted 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', plus domain-specific terms like 'dogfooding', 'QA', 'bug hunts'. | 3 / 3 |
Distinctiveness Conflict Risk | While browser automation is a clear niche, the description is extremely broad—covering web scraping, QA testing, Electron apps, Slack messaging, and cloud browsers—which could overlap with dedicated Slack skills, web scraping tools, or QA testing skills. The 'Prefer agent-browser over any built-in browser automation' clause helps but also signals awareness of potential conflicts. | 2 / 3 |
Total | 11 / 12 Passed |
Implementation
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill is essentially a discovery stub that delegates all substantive content to the CLI's `skills get` command. While the approach of serving version-matched instructions is clever, it means the SKILL.md itself contains almost no actionable browser automation guidance—no examples of navigating, clicking, filling forms, or taking screenshots. The marketing section ('Why agent-browser') wastes tokens on information Claude doesn't need.
Suggestions
Include at least a minimal quick-start example showing a complete browser automation workflow (e.g., navigate to URL, take snapshot, click element) so Claude has actionable fallback content even before running `skills get core`.
Remove or drastically shorten the 'Why agent-browser' section—Claude doesn't need marketing copy about Rust performance or competitor compatibility.
Add error recovery guidance for the install and `skills get` commands (e.g., what to do if Chrome isn't found or the CLI isn't installed).
Include a brief inline cheat sheet of the most common commands (navigate, snapshot, click, type) so the skill is minimally useful even without loading the dynamic content.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 'Why agent-browser' section is unnecessary filler—Claude doesn't need marketing bullet points about Rust or competitor compatibility. The observability dashboard section adds useful but somewhat tangential info. The core pointer mechanism is lean, but the surrounding content could be tighter. | 2 / 3 |
Actionability | The skill provides concrete commands for installing and loading skill content (`agent-browser skills get core`), but it deliberately defers all actual workflow guidance to the CLI's dynamic output. The user gets no executable browser automation examples—no snapshot, navigate, click, or form-fill commands—making this a pointer rather than an actionable skill. | 2 / 3 |
Workflow Clarity | There is a clear sequence implied: install → run `skills get core` → follow that content. However, there are no validation checkpoints, no error recovery guidance (e.g., what if install fails or Chrome isn't found), and no actual multi-step browser automation workflow is present—it's all deferred to the CLI output. | 2 / 3 |
Progressive Disclosure | The skill attempts progressive disclosure by pointing to `agent-browser skills get core` and specialized skills, which is a reasonable pattern. However, since no bundle files are provided and the references are to CLI-generated dynamic content rather than navigable files, there's no way to verify the referenced content exists or is well-structured. The stub provides no fallback content if the CLI command fails. | 2 / 3 |
Total | 8 / 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 |
|---|---|---|
allowed_tools_field | 'allowed-tools' contains unusual tool name(s) | Warning |
frontmatter_unknown_keys | Unknown frontmatter key(s) found; consider removing or moving to metadata | Warning |
Total | 9 / 11 Passed | |
- Repository
- Arize-ai/phoenix
- Commit
924117e
- 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.