computer-use-agents

The fundamental architecture of computer use agents: observe screen, reason about next action, execute action, repeat. This loop integrates vision models with action execution through an iterative pipeline.

Quality

50%

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 ./plugins/AI-Agents-Safe-Coding-Skills/skills/computer-use-agents/SKILL.md

The canonical home for this skill is computer-use-agents in sickn33/agentic-awesome-skills

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 body offers substantive concrete material but is undermined by truncated, non-executable code samples, a broken placeholder 'Sharp Edges' table, dated version claims, and a monolithic structure that inlines content which should live in reference files. Validation steps for risky automation are absent.

Suggestions

Complete or remove the truncated Python code blocks so every example is copy-paste executable.

Move the Dockerfile, docker-compose, and full agent implementations into reference files under references/ and link to them one level deep from SKILL.md.

Replace the placeholder 'Sharp Edges' table with real issue names and add explicit validation/verification checkpoints (e.g., confirm sandbox isolation before running, verify action results before proceeding).

Dimension	Reasoning	Score
Conciseness	Mostly structured but padded with a large inline Dockerfile and docker-compose, explains the perception-reasoning-action loop Claude already knows, and embeds time-sensitive version strings ('claude-sonnet-4-20250514', 'computer_20251124', 'Opus 4.5 is now the best') outside any deprecated section.	3 / 5
Actionability	Provides real concrete guidance (a complete Dockerfile and docker-compose), but the Python examples are truncated mid-line ('scrolled {dir', 'from dataclasses im', an unfinished _handle_computer_action), so they are not executable as written.	3 / 5
Workflow Clarity	The loop and isolation requirements are sequenced, but this is destructive/risky desktop automation with no validation checkpoints, and the 'Sharp Edges' table uses placeholder 'Issue' labels with no real problems, so workflow clarity is capped at 3.	3 / 5
Progressive Disclosure	No bundle files exist; large artifacts (Dockerfile, docker-compose, full agent classes) that belong in separate reference files are inlined into SKILL.md with no one-level-deep references, and at ~320 lines the simple-skill exception does not apply.	2 / 5
	Total	11 / 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 clearly states the architectural pattern of computer use agents but omits any explicit 'when to use' trigger guidance and relies on limited natural keywords. It is a competent but incomplete description that would benefit from trigger phrases.

Suggestions

Add an explicit 'Use when...' clause listing concrete trigger scenarios (e.g., building desktop automation, controlling a GUI, driving a browser/OS).

Broaden natural trigger terms with synonyms users actually say, such as 'desktop automation', 'GUI control', or 'screen interaction'.

Reframe the 'what' around deliverables (e.g., 'drives mouse/keyboard to operate a desktop') rather than only the abstract loop.

Dimension	Reasoning	Score
Specificity	Names the domain ('computer use agents') and concrete loop phases ('observe screen, reason about next action, execute action, repeat'), but these describe architecture rather than user-facing deliverable actions, matching the 'names domain and 1-2 concrete actions' anchor.	3 / 5
Completeness	Provides a clear 'what' (the perception-reasoning-action loop) but no 'Use when...' trigger clause, so per the rubric completeness is capped at 3.	3 / 5
Trigger Term Quality	Only natural term is 'computer use agents'; the rest ('vision models', 'iterative pipeline') is technical jargon, and common user phrases like 'control my computer' or 'automate desktop' are missing.	3 / 5
Distinctiveness Conflict Risk	'Computer use agents' is a fairly distinct niche with minor overlap risk against general agent skills, fitting the 'mostly distinct' anchor; not a 5 because the description frames a generic agent loop.	4 / 5
	Total	13 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: administrakt0r/AI-Agents-Safe-Coding-Skills
Path: plugins/AI-Agents-Safe-Coding-Skills/skills/computer-use-agents/SKILL.md
Commit: 57c135b

Reviewed: 2 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.