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.

Quality

56%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Fix and improve this skill with Tessl

tessl review fix ./skills/computer-use-agents/SKILL.md

Quality

Content

46%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 highly actionable with extensive executable code and useful sharp-edges guidance, but it is a monolithic ~2157-line file that violates progressive disclosure and carries notable verbosity. Splitting implementations into reference files and trimming explanatory prose would substantially improve it.

Suggestions

Move the large class implementations (ComputerUseAgent, SandboxedAgent, AnthropicComputerUse, BrowserUseAgent, ConfirmationGate, ActionLogger, CostTracker) and the Dockerfile/docker-compose into separate reference files under references/, keeping only concise overviews and key snippets in SKILL.md.

Trim explanatory prose that restates what Claude already knows (e.g., why humans click off-center, generic cost-growth narration) to reduce token load.

Weave explicit validate→fix→retry checkpoints into the destructive/batch workflows (sandboxing, bash tool, drag/scroll) rather than listing validation as a separate catalog.

Dimension	Reasoning	Score
Conciseness	The ~2157-line body inlines six-plus full multi-class implementations plus Dockerfiles and explanatory prose that clearly belong in separate reference files, making it noticeably verbose with several padded sections.	2 / 5
Actionability	Provides concrete, mostly copy-paste-ready implementations (ComputerUseAgent, SandboxedAgent, AnthropicComputerUse, BrowserUseAgent, ConfirmationGate, ActionLogger, CostTracker) with only minor gaps (placeholder `pass`, undefined `task_complete`/`vision_agent`).	4 / 5
Workflow Clarity	The perception-reasoning-action loop and sandboxing requirements are numbered with a feedback step and a dedicated Validation Checks section, but the skill is a pattern catalog whose destructive/batch workflows lack inline validate→fix→retry checkpoints.	3 / 5
Progressive Disclosure	No bundle files exist (references/, scripts/, assets/ empty) and all large implementations are inlined into one monolithic SKILL.md with only section headers, matching the minimal-structure anchor.	2 / 5
	Total	11 / 20 Passed

Description

66%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 clearly conveys what the skill builds and names concrete capabilities with good trigger keywords, but it omits an explicit "Use when..." trigger clause in the frontmatter. Adding trigger phrases and a few more natural synonyms would lift completeness and trigger-term quality.

Suggestions

Add an explicit "Use when..." clause naming the trigger situations (e.g., desktop automation, GUI automation, vision-based agents) to raise completeness above 3.

Include more natural synonyms users say ("GUI automation", "RPA", "desktop automation", "screen control") to broaden trigger-term coverage.

Tighten the phrasing into a single concise sentence to avoid slight padding while preserving the concrete action list.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions ("viewing screens, moving cursors, clicking buttons, and typing text") plus distinct sub-focuses across Anthropic/OpenAI/open-source, falling just short of the fully comprehensive 5-anchor example.	4 / 5
Completeness	Clearly answers the "what" (building computer-use agents) but the frontmatter description lacks any explicit "Use when..." trigger guidance, capping completeness at 3 per the missing-trigger guideline.	3 / 5
Trigger Term Quality	Good natural keyword coverage ("computer use", "Operator/CUA", "vision-based", "AI agents") but omits common variations like "GUI automation", "RPA", or "desktop automation" that the body itself uses.	4 / 5
Distinctiveness Conflict Risk	"Computer Use Agents" is a distinct niche (Anthropic Computer Use, Operator, vision-based control) with minimal conflict risk, though it slightly overlaps generic browser-automation/automation skills.	4 / 5
	Total	15 / 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
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	14 / 16 Passed

Repository: sickn33/antigravity-awesome-skills
Path: skills/computer-use-agents/SKILL.md
Commit: 1f67c44

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.