agent-browser

Browse the web for any task — research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages. Use whenever a browser would be useful, not just when the user explicitly asks.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

78%Scale 1-5

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

This is a well-crafted, highly actionable skill with concrete, executable CLI commands covering a broad range of browser automation tasks. Its main strength is actionability — every command is copy-paste ready with clear syntax. Its main weakness is that the extensive command reference would benefit from being split into a concise overview and a separate reference file, and the workflow could include light error-handling guidance.

Suggestions

Split the detailed command sections (cookies, storage, JavaScript, semantic locators, authentication) into a separate REFERENCE.md and keep SKILL.md as a concise overview with the core workflow and most common commands.

Add brief error-handling guidance to the core workflow, e.g., 'If snapshot returns no interactive elements, try `snapshot` without `-i` flag or wait for page load with `wait --load networkidle`'.

Dimension	Reasoning	Score
Conciseness	The content is efficient and avoids explaining what a browser is or how automation works. It's mostly lean, though the sheer breadth of command listings (cookies, storage, JavaScript, semantic locators, authentication) makes it slightly longer than necessary for a skill file — some of these could be in a reference file. Minor redundancy between 'fill' and 'type' explanations.	4 / 5
Actionability	Every command is concrete, copy-paste ready with exact CLI syntax. The examples for form submission and data extraction show complete workflows with realistic commands. No pseudocode or vague instructions — everything is directly executable.	5 / 5
Workflow Clarity	The core workflow (navigate → snapshot → interact → re-snapshot) is clearly sequenced and the form submission example demonstrates it well. However, there are no explicit validation/error-handling steps — e.g., what to do if a snapshot returns no interactive elements, or if a click doesn't navigate as expected. Since this isn't a destructive/batch operation, the missing feedback loops don't cap the score at 3, but they prevent a 5.	4 / 5
Progressive Disclosure	All content is inlined in a single file with no references to external files. The comprehensive command reference (navigation, snapshots, interactions, get, screenshots, wait, semantic locators, auth, cookies, storage, JavaScript) would benefit from being split into a quick-start overview and a separate REFERENCE.md. The structure within the file is decent with clear section headers, but the volume of content makes it more of a reference manual than an overview.	3 / 5
	Total	16 / 20 Passed

Description

88%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.

This is a strong skill description that clearly communicates both what the skill does and when to use it. The enumeration of specific actions is comprehensive, and the 'when' clause is notably well-crafted by encouraging proactive use. Minor improvements could include additional natural trigger terms and synonyms to further strengthen keyword matching.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete actions: 'research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages.' This is comprehensive coverage of web browsing capabilities.	5 / 5
Completeness	Clearly answers both 'what' (browse the web for research, reading, interacting, forms, screenshots, data extraction, testing) and 'when' ('Use whenever a browser would be useful, not just when the user explicitly asks'). The 'when' clause is explicit and even broadens the trigger beyond explicit requests.	5 / 5
Trigger Term Quality	Includes good natural keywords like 'browse the web', 'research', 'articles', 'forms', 'screenshots', 'web pages', and 'web apps'. Missing some natural synonyms like 'website', 'URL', 'navigate', 'search', or 'internet' that users might say.	4 / 5
Distinctiveness Conflict Risk	The web browsing domain is fairly distinct from most other skills. However, 'research topics' and 'extract data' could overlap with research or data extraction skills that don't involve a browser. The 'browser' framing helps distinguish it, but there's minor overlap risk.	4 / 5
	Total	18 / 20 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 — 11 / 11 Passed

Validation for skill structure

No warnings or errors.

Repository: jbaruch/nanoclaw-telegram
Path: container/skills/agent-browser/SKILL.md
Commit: 7454673

Reviewed: 1 day 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.