agent-browser

Browser automation through the repo agent-browser CLI. Explicit helper for navigation, forms, screenshots, scraping, and web-app checks. Prefer Browser Use or Playwright when available. Do NOT load for: sharing URLs, embedding links, or editing screenshot files.

Quality

75%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

High

Do not use without reviewing

Fix and improve this skill with Tessl

tessl review fix ./opencode/skills/agent-browser/SKILL.md

Quality

Content

75%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 skill body with executable examples and real reference files, held back by redundancy between the request-classification and quick-reference sections and by reference-style content kept inline. Workflow sequencing is clear with a verification step but lacks an explicit error-recovery loop.

Suggestions

Collapse the Step 1 request-classification table into the Quick Reference section (or vice versa) to remove the duplicated command listings and recover tokens.

Add an explicit feedback loop after Step 3, e.g. 'If the expected element is missing or the action did not take effect, re-run snapshot -i -c and retry with the new ref'.

Move the five Quick Reference command tables (basic ops, navigation, info, wait, debug) into a references/cheat-sheet file, keeping only the core snapshot workflow inline in SKILL.md.

Dimension	Reasoning	Score
Conciseness	Content is mostly efficient command tables with no concept over-explanation, but the Step 1 request-classification table duplicates the Quick Reference tables and the trigger-phrases section repeats what belongs in the description, so it could be tightened.	3 / 5
Actionability	Fully executable, copy-paste-ready agent-browser commands throughout, including a realistic end-to-end example with element refs (@e2, @e3) and sample snapshot output covering the common navigation/form-flow case.	5 / 5
Workflow Clarity	Steps 0–3 are clearly sequenced and Step 3 ('結果の確認') provides verification via snapshot/url/screenshot, but there is no explicit 'if the action failed, re-snapshot and retry' feedback loop, leaving a minor validation gap below anchor 5.	4 / 5
Progressive Disclosure	Two one-level-deep references (browser-automation.md, ai-snapshot-workflow.md) are well-signaled via ${CLAUDE_SKILL_DIR} paths and both exist, but five inline quick-reference tables plus session-management and MCP-comparison content is substantial reference material that could be delegated, a minor organization gap.	4 / 5
	Total	16 / 20 Passed

Description

75%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 specific, well-scoped description with strong distinctiveness and concrete capability listing, weakened only by the absence of a positive 'Use when...' trigger clause. The negative boundary is effective but leaves the positive trigger half implied.

Suggestions

Add a positive trigger clause such as 'Use when the user asks to open a page, click elements, fill forms, take screenshots, scrape content, or verify a web UI' to satisfy the 'when' half of completeness.

Include a few natural phrasal trigger terms users actually say (e.g. 'open this page', 'click the button', 'fill in the form') alongside the current domain terms.

Consider naming the CLI's primary trigger verb up front (e.g. 'Drive the agent-browser CLI to...') so the action voice reads more naturally in third person.

Dimension	Reasoning	Score
Specificity	Lists five concrete actions — 'navigation, forms, screenshots, scraping, and web-app checks' — giving comprehensive coverage of the skill's capabilities, matching the anchor for multiple specific concrete actions.	5 / 5
Completeness	The 'what' is clear ('Browser automation through the repo agent-browser CLI' plus listed actions), but there is no positive 'Use when...' clause — 'when' is only a negative boundary ('Do NOT load for: sharing URLs, embedding links, or editing screenshot files') plus implied intent, which caps completeness at 3 per the rubric guideline.	3 / 5
Trigger Term Quality	Terms like 'forms', 'screenshots', 'scraping', and 'web-app checks' are fairly natural, but common phrasal variations users actually say ('open this page', 'click', 'fill the form') and any file extensions are absent, leaving a few natural terms missing.	4 / 5
Distinctiveness Conflict Risk	The explicit negative boundary ('Do NOT load for: sharing URLs, embedding links, or editing screenshot files') plus alternative-tool guidance ('Prefer Browser Use or Playwright when available') carves out a clear niche with minimal conflict risk against adjacent skills.	5 / 5
	Total	17 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
relative_links	Relative link issues: 2 missing, 2 deeper-than-1-level	Warning

	Total	15 / 16 Passed

Repository: Chachamaru127/claude-code-harness
Path: opencode/skills/agent-browser/SKILL.md
Commit: 7d8deba

Reviewed: about 15 hours 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.