computer-use-agents
Build AI agents that interact with computers like humans do - viewing screens, moving cursors, clicking buttons, and typing text. Covers Anthropic's Computer Use, OpenAI's Operator/CUA, and open-source alternatives.
49
55%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Risky
Do not use without reviewing
Optimize this skill with Tessl
npx tessl skill review --optimize ./skills/computer-use-agents/SKILL.mdQuality
Discovery
67%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 is strong in specificity and distinctiveness, clearly naming concrete actions and specific platforms/products. Its main weaknesses are the lack of an explicit 'Use when...' clause and missing some natural trigger terms users might employ when seeking help with GUI automation or computer-use agents.
Suggestions
Add an explicit 'Use when...' clause, e.g., 'Use when the user asks about building computer-use agents, GUI automation, screen interaction, or browser control with AI.'
Include additional natural trigger terms like 'browser automation', 'GUI automation', 'desktop automation', 'RPA', 'screen scraping', or 'web agent' to improve keyword coverage.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: 'viewing screens, moving cursors, clicking buttons, and typing text' along with specific platforms covered (Anthropic's Computer Use, OpenAI's Operator/CUA, open-source alternatives). | 3 / 3 |
Completeness | Clearly answers 'what' (build AI agents that interact with computers via screen/cursor/click/type, covering specific platforms), but lacks an explicit 'Use when...' clause or equivalent trigger guidance for when Claude should select this skill. | 2 / 3 |
Trigger Term Quality | Includes some good natural keywords like 'AI agents', 'computer use', 'clicking buttons', 'typing text', and specific product names. However, it misses common user terms like 'browser automation', 'GUI automation', 'screen control', 'RPA', or 'desktop automation' that users might naturally say. | 2 / 3 |
Distinctiveness Conflict Risk | The description carves out a very clear niche around computer-use AI agents with specific product names (Anthropic's Computer Use, OpenAI's Operator/CUA), making it highly distinguishable from general coding, automation, or other AI skills. | 3 / 3 |
Total | 10 / 12 Passed |
Implementation
42%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill provides highly actionable, executable code covering computer use agents comprehensively, which is its primary strength. However, it is severely over-long and monolithic - at ~1500 lines with no external file references, it would consume enormous context window space while repeating security concepts across multiple sections. The workflow clarity is adequate but lacks explicit validation checkpoints for these security-critical operations.
Suggestions
Split into multiple files: keep SKILL.md as a concise overview (~100-150 lines) with quick-start code, and move detailed implementations (AnthropicComputerUse class, BrowserUseAgent, ActionLogger, etc.) into referenced files like ANTHROPIC_IMPL.md, BROWSER_USE.md, LOGGING.md, SECURITY.md
Remove explanations of concepts Claude already knows - e.g., why sandboxing matters, what Docker capabilities are, why vision tokens are expensive, what prompt injection is. Replace with terse directives.
Add explicit validation checkpoints to workflows - e.g., after sandbox creation verify isolation with a test command, after agent action verify the expected UI state changed, after container setup verify non-root and capability drops
Consolidate the repeated security advice (sandboxing appears in Patterns, Sharp Edges, and Validation Checks) into a single authoritative section with cross-references
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | This skill is extremely verbose at ~1500+ lines. It explains concepts Claude already knows (what a perception-reasoning-action loop is, why sandboxing matters, what Docker does), includes massive code blocks that could be condensed, and repeats security advice across multiple sections. The anti-patterns for each section, the sharp edges, and the validation checks all contain significant redundancy. | 1 / 3 |
Actionability | The skill provides fully executable, copy-paste ready code throughout - complete Python classes, Docker configurations, docker-compose files, and concrete command-line examples. Every pattern includes working implementations with specific libraries, methods, and configurations. | 3 / 3 |
Workflow Clarity | The perception-reasoning-action loop is clearly sequenced, and the sandbox setup has logical steps. However, there are no explicit validation checkpoints in the workflows - for example, the sandbox setup doesn't verify the container is properly isolated before running the agent, and the main agent loop lacks verification that actions actually succeeded before proceeding. For destructive/security-critical operations like these, explicit validation steps are needed. | 2 / 3 |
Progressive Disclosure | This is a monolithic wall of text with no references to external files despite being ~1500 lines. The Anthropic implementation, browser-use pattern, confirmation gate, action logging, and all sharp edges are inlined. Content like the full ActionLogger class, CostTracker, and ContextManager should be in separate reference files. There are no bundle files to support progressive disclosure. | 1 / 3 |
Total | 7 / 12 Passed |
Validation
81%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 9 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
skill_md_line_count | SKILL.md is long (2167 lines); consider splitting into references/ and linking | Warning |
frontmatter_unknown_keys | Unknown frontmatter key(s) found; consider removing or moving to metadata | Warning |
Total | 9 / 11 Passed | |
- Repository
- sickn33/antigravity-awesome-skills
- Commit
f5dc9e3
- 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.