CtrlK
BlogDocsLog inGet started
Tessl Logo

chrome-automation

网页开发工程师与运营专员在需要自动化现有 Chrome 浏览器时必用!通过 CDP 协议无缝接管当前浏览器,自动配置环境并实时可视化执行任务。完美保留登录状态,一键完成复杂网页自动化流程,让繁琐操作自动化。

52

Quality

59%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./skills/chrome-automation/chrome-automation/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

56%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 body is highly actionable and the runtime workflow is clearly sequenced with real, present scripts — its main strengths. Its weaknesses are severe redundancy (installation and Rust troubleshooting each covered 2-3 times), inlined content that duplicates the reference files the bundle already ships, and machine-specific hardcoded paths that undermine copy-paste executability. Deduplicating into references and parameterizing the install path would move this from adequate to strong.

Suggestions

State the installation story exactly once: keep the auto-install invocation plus a verify step in SKILL.md and move the entire Rust/toolchain troubleshooting (Issues 1-4 and their duplicates in 'Troubleshooting') into references/troubleshooting.md, which already exists for this purpose.

Replace every hardcoded `/Users/jyxc-dz-0100272/Documents/agent-browser` path with a parameterized variable (e.g. `export AB_HOME="$HOME/Documents/agent-browser"` defined once) and remove the machine-specific pointer to the agent-browser SKILL.md in favor of the bundle's own references/commands.md.

Replace the inline 'Essential Commands' list with a pointer to references/commands.md (a few flagship examples inline, the full reference linked), and cut the duplicated start-chrome/auto-install instructions from Quick Start to a single line each.

DimensionReasoningScore

Conciseness

The ~670-line body repeats the same material multiple times: the auto-install script invocation appears in the CRITICAL section, Quick Start, and Installation & Setup; the Rust PATH/toolchain fixes appear verbatim in both 'Common Rust Installation Issues' and again in 'Troubleshooting'; the start-chrome instructions appear twice with identical explanations. This is noticeably verbose with several padded sections (anchor 2), not merely trimmable (anchor 3).

2 / 5

Actionability

Commands are concrete and copy-paste ready (auto-install/start-chrome script invocations, `agent-browser --cdp 9222 open/snapshot -i/fill @e1/screenshot` patterns, platform-specific PowerShell equivalents) and the referenced scripts (auto-install-mac.sh, start-chrome-mac.sh, etc.) actually exist in the bundle. It falls short of 5 because core operation patterns hardcode one user's home directory — `cd /Users/jyxc-dz-0100272/Documents/agent-browser` — and point to a full command reference at that same machine-specific absolute path, so the guidance is not executable as written on any other machine.

4 / 5

Workflow Clarity

The Core Workflow (Connect → Navigate → Snapshot → Interact → Screenshot → Report) is clearly sequenced, installation ends with an explicit verification step ('Verifies installation success', `--version` check), and the Real-time Feedback Protocol plus 'Only proceed AFTER successful installation' and 'Check URL' best practices act as checkpoints. It is not 5 because some checkpoints are implicit rather than scripted into the workflow (e.g., no explicit check that the CDP connection succeeded before navigating, and re-snapshot guidance lives only in a best-practices list).

4 / 5

Progressive Disclosure

The bundle does contain references/ (setup-mac.md, setup-windows.md, troubleshooting.md, commands.md) and the body signals three of them ('For detailed troubleshooting, see: references/setup-mac.md...'), all one level deep. But the split is half-done: an inline 'Essential Commands' section duplicates references/commands.md, which is never linked, and the 100+ lines of Rust installation troubleshooting inline in SKILL.md duplicate what references/troubleshooting.md and setup-*.md are for. This matches anchor 3 (references present and signaled, but content that should be separate is inline).

3 / 5

Total

13

/

20

Passed

Description

62%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 real, differentiated capability (CDP takeover of the user's Chrome with preserved logins) and includes an explicit when-clause, but it reads as promotional copy: buzzwords ('seamless', 'perfect', 'one-click', 'must-use!') crowd out specific actions and natural trigger variations. Rewriting it in plain third-person operational language with concrete verbs and common trigger synonyms would lift every dimension.

Suggestions

Replace marketing language ('无缝', '完美保留', '一键', '必用!') with concrete third-person actions, e.g. 'Connects to the user's running Chrome via CDP, preserves login state, and automates navigation, form filling, and element interaction with screenshots after each step.'

Add natural trigger terms and synonyms users would actually say: 'browser automation', '网页自动化', '自动操作网页', 'browser control', 'automate login-required sites', so the skill surfaces for those phrasings.

Make the 'when' clause differentiate from generic web automation, e.g. 'Use when the user wants to automate their existing Chrome browser with logged-in sessions visible in real time.'

DimensionReasoningScore

Specificity

The description names the domain (automating an existing Chrome browser) and a few concrete capabilities ('通过 CDP 协议...接管当前浏览器', '保留登录状态', '实时可视化执行任务'), but these are buried in marketing fluff ('无缝', '完美', '一键', '必用!') and the core action '执行任务'/'完成复杂网页自动化流程' is generic. It matches anchor 3 (domain plus 1-2 concrete actions, not comprehensive); it is below 4 because buzzwords replace a fuller list of specific actions like navigate, fill forms, screenshot, or extract data.

3 / 5

Completeness

Both 'what' and 'when' are explicitly present: '在需要自动化现有 Chrome 浏览器时必用' is an explicit trigger clause, and the 'what' is stated as taking over the current browser via CDP, preserving login state, and executing automation with real-time visibility. It does not reach 5 because the 'when' clause is largely a restatement of the 'what' (use it when you need browser automation) rather than concrete, differentiated trigger phrases, and the 'what' is padded with over-claims instead of comprehensive concrete actions.

4 / 5

Trigger Term Quality

Relevant keywords exist — 'Chrome', 'CDP 协议', '自动化现有 Chrome 浏览器', '网页自动化' — which a user needing this skill might naturally say. However common variations are missing: no '浏览器自动化' (browser automation), '网页操作/自动操作', '爬取/scrape', 'browser control', or tool/protocol synonyms beyond CDP. This fits anchor 3 (some relevant keywords but missing common variations or synonyms) rather than 4, which requires only a few missing natural terms.

3 / 5

Distinctiveness Conflict Risk

The niche — controlling the user's existing Chrome instance via CDP while preserving login state — is fairly distinct and unlikely to fire for unrelated skills. It stays at 4 rather than 5 because the closing phrase '复杂网页自动化流程' / '繁琐操作自动化' is broad enough to overlap with general web-automation or scraping skills, and no tool name (agent-browser) or sharper trigger disambiguates it.

4 / 5

Total

14

/

20

Passed

Validation

87%

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

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (671 lines); consider splitting into references/ and linking

Warning

referenced_paths_exist

Referenced path issues: 4 missing

Warning

Total

14

/

16

Passed

Repository
anbeime/skill
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.