gemini-computer-use

Build and run Gemini 2.5 Computer Use browser-control agents with Playwright. Use when a user wants to automate web browser tasks via the Gemini Computer Use model, needs an agent loop (screenshot → function_call → action → function_response), or asks to integrate safety confirmation for risky UI actions.

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Advisory

Suggest reviewing before use

Quality

Content

85%

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

A well-structured, lean skill body with a clearly sequenced agent loop and excellent reference signaling; its one real defect is the missing env.example file, which breaks the first quick-start command.

Suggestions

Add the missing env.example file to the bundle (or inline the required environment variables, e.g. the API key variable name) so the first quick-start command 'cp env.example env.sh' is copy-paste ready.

Specify which environment variable holds the API key in the Quick start step so users know what to set after copying env.example.

Optionally cross-check that the script's --prompt, --start-url, and --turn-limit flags exactly match the documented run command.

Dimension	Reasoning	Score
Conciseness	The body is lean and assumes Claude's competence — no explanation of what Playwright or a screenshot is, just tight command blocks and specific operational guidance, with every section earning its place.	3 / 3
Actionability	Most guidance is concrete and executable (venv/pip/playwright install, the run command with real --prompt/--start-url/--turn-limit flags, --exclude, viewport 1440x900), but the very first quick-start command 'cp env.example env.sh' references an env.example file that is not present in the bundle, so it is not copy-paste ready as written.	2 / 3
Workflow Clarity	The six-step agent loop is clearly sequenced with an explicit validation checkpoint (step 4: 'If a safety_decision is require_confirmation, prompt the user before executing') for the risky UI actions this skill performs, plus an explicit termination condition and an inherent screenshot feedback loop.	3 / 3
Progressive Disclosure	SKILL.md is a concise overview that pushes detail to one-level-deep, clearly signaled references in a Resources section ('scripts/computer_use_agent.py', 'references/google-computer-use.md'), both of which exist in the bundle; content is appropriately split rather than inlined.	3 / 3
	Total	11 / 12 Passed

Description

85%

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 strong, specific description with explicit 'Use when' triggers and a clear niche; its main weakness is trigger phrasing that mixes user-natural terms with internal API-mechanism jargon.

Suggestions

Replace internal mechanism terms like 'function_call / function_response' and 'safety_decision' in the trigger clause with user-natural phrasing such as 'automate browser clicks and form fills' or 'control a web page'.

Add common natural variations users would actually say, e.g. 'control the browser', 'click buttons and fill forms', or 'scrape a web page'.

Keep the 'Use when' structure but move the function_call/function_response detail into the body so the description stays trigger-focused.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'Build and run Gemini 2.5 Computer Use browser-control agents with Playwright', 'automate web browser tasks', and 'integrate safety confirmation' — rather than vague language.	3 / 3
Completeness	Clearly states what it does ('Build and run ... browser-control agents with Playwright') and when to use it via an explicit 'Use when a user wants ... or asks to ...' clause with multiple triggers.	3 / 3
Trigger Term Quality	Includes some natural terms ('automate web browser tasks', 'Gemini Computer Use') but leans on internal mechanism jargon ('agent loop (screenshot → function_call → action → function_response)', 'safety confirmation for risky UI actions') that users would not naturally say, and misses common variations like 'control the browser' or 'click and fill forms'.	2 / 3
Distinctiveness Conflict Risk	Targets a clear niche (Gemini 2.5 Computer Use + Playwright browser control) with distinct, model-specific triggers, making overlap with other skills unlikely.	3 / 3
	Total	11 / 12 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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: am-will/codex-skills
Commit: e343715

Reviewed: 14 days 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.