agent-browser

Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection

Quality

68%

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

Fix and improve this skill with Tessl

tessl review fix ./configs/microservice/bff-service/configs/agent-skills/clawhub/agent-browser/SKILL.md

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.

The content is a strong, executable command reference: fully copy-paste-ready commands, a clear core workflow with re-snapshot discipline, and two worked end-to-end examples. Its weaknesses are modest — implicit rather than explicit validation checkpoints, and a monolithic structure where the full command catalog could live in a separate reference file.

Dimension	Reasoning	Score
Conciseness	The body is a lean command catalog ('agent-browser snapshot -i --json # Interactive elements, JSON output') that assumes Claude's competence and explains no known concepts, matching 'efficient; minor instances of over-explanation that could be trimmed' — the 'Why Use This Over Built-in Browser Tool' bullet list and Credits section are the only minor padding.	4 / 5
Actionability	Every section is copy-paste-ready executable commands, and the worked examples ('Example: Search and Extract', 'Example: Multi-Session Testing') cover the common cases ('agent-browser fill @e1 "AI agents"' → 'press Enter' → 'wait --load networkidle'), matching the 'fully executable; copy-paste ready' anchor exactly.	5 / 5
Workflow Clarity	'Core Workflow' lays out a clear numbered navigate → snapshot → parse refs → interact → re-snapshot sequence with stability waits, matching 'clear sequence with most checkpoints present; minor validation gaps' — explicit validation of the JSON success/refs output before interacting is only implied ('Parse refs from JSON'), not stated as a checkpoint, so it falls short of the anchor 5 feedback-loop pattern.	4 / 5
Progressive Disclosure	The single-file body is well organized into clearly headed sections ('## Key Commands', '### Snapshot (Always use -i --json)', '## Example: Search and Extract') with no bundle files and no buried or nested references, matching 'good structure; most content is appropriately placed; minor organization gaps' — a ~150-line inline command catalog is borderline content that could be split one level deeper into a reference file, which keeps it below the well-signaled multi-file anchor at 5.	4 / 5
	Total	17 / 20 Passed

Description

53%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.

The description communicates a clear and fairly distinctive 'what' with concrete technical mechanisms, but it omits any 'when to use' trigger guidance and lists no concrete actions, limiting both completeness and trigger coverage. It reads like a tool characterization rather than a capability-plus-trigger statement.

Suggestions

Add an explicit trigger clause, e.g. 'Use when automating web pages, filling forms, scraping content, or testing SPAs in a headless browser' — the missing 'Use when...' currently caps completeness at 3.

Enumerate 2-4 concrete capabilities (e.g. 'click and fill elements by ref, extract text/HTML/attributes, wait for network idle, manage isolated sessions') so specificity moves from 1-2 named mechanisms to a several-action list.

Include natural synonyms users actually say — 'web scraping', 'screenshot', 'navigate', 'browser testing', 'Playwright alternative' — to improve trigger-term coverage.

Dimension	Reasoning	Score
Specificity	Names the domain ('Headless browser automation CLI') and 1-2 concrete mechanisms ('accessibility tree snapshots', 'ref-based element selection') but never lists the actual actions it performs (click, fill, extract text, screenshot), so it matches the 'names domain and 1-2 concrete actions, but not comprehensive' anchor rather than the several-actions anchor at 4.	3 / 5
Completeness	The 'what' is clear (a browser automation CLI using accessibility-tree snapshots and refs), but there is no 'Use when...' clause or any equivalent explicit trigger guidance, which per the judging guidelines caps completeness at 3.	3 / 5
Trigger Term Quality	'browser automation' and 'headless browser' are relevant keywords, but common natural phrases users would say — 'scrape', 'screenshot a page', 'navigate a website', 'fill a form', 'Playwright' — are absent, matching the 'some relevant keywords but missing common variations or synonyms' anchor.	3 / 5
Distinctiveness Conflict Risk	The description carves out a fairly distinct niche (agent-optimized CLI with accessibility-tree/ref-based selection) so it is 'mostly distinct; minor overlap risk with closely related skills' — generic browser-automation or web-scraping skills could still compete for the same triggers, keeping it below the clear-niche anchor at 5.	4 / 5
	Total	13 / 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: UnicomAI/wanwu
Path: configs/microservice/bff-service/configs/agent-skills/clawhub/agent-browser/SKILL.md
Commit: a16301c

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.