cua-driver

Drive a native macOS app via the cua-driver MCP server or CLI — snapshot its AX tree, click/type/scroll by element_index, verify via re-snapshot. Use when the user asks you to operate, drive, automate, or perform a GUI task in a real macOS application on the host (e.g. "open a file in TextEdit", "navigate to /Applications in Finder", "click the Save button in Numbers").

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

77%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 with an excellent sequenced workflow and explicit validation loops, but it is noticeably verbose with repetitive no-foreground-contract restatements, and its progressive-disclosure references point to bundle files that are not present.

Suggestions

Consolidate the no-foreground-contract material into one canonical section and have later references link back to it instead of restating the prohibition and the LaunchServices rationale each time — this would cut hundreds of lines.

Add the referenced bundle files (WEB_APPS.md, RECORDING.md, SCREENSHOT.md) to a references/ directory, or remove the links and inline the minimum needed content, so the signaled navigation actually resolves.

Trim mechanics Claude can be assumed to know (LaunchServices routing, HID event-source filtering) to the specific behavioral facts that affect tool choice, keeping the token budget focused on what is non-obvious.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and rich with genuine domain knowledge, but at ~940 lines it repeats the no-foreground contract across several sections (Prerequisites, Defaults, self-check, carve-out) and over-explains LaunchServices mechanics Claude could be assumed to follow.	3 / 5
Actionability	Fully executable throughout — copy-paste CLI invocations ('cua-driver launch_app {…}', 'get_window_state … --image-out /tmp/shot.png'), a complete canonical workflow, an intent→tool dispatch table, and a runnable PIL annotation snippet cover the common cases.	5 / 5
Workflow Clarity	The canonical loop (launch_app → get_window_state → act → get_window_state → verify) is explicitly sequenced with validation checkpoints on both sides of every action, a self-check pattern, and a feedback-loop error-recovery table; destructive/batch actions are gated on user intent.	5 / 5
Progressive Disclosure	The body is well-structured and points one-level-deep to clearly signaled files (WEB_APPS.md, RECORDING.md, SCREENSHOT.md), but none of those referenced bundle files actually exist on disk — a real organization gap that leaves navigation broken.	3 / 5
	Total	16 / 20 Passed

Description

92%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 is exemplary: third-person voice, concrete actions, explicit 'Use when' trigger guidance with natural example phrases, and a clearly distinct niche. Its only minor gap is missing synonym/extension-style keyword enumeration, which keeps trigger_term_quality just below a 5.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete actions — 'snapshot its AX tree, click/type/scroll by element_index, verify via re-snapshot' — with comprehensive coverage of the skill's capability surface.	5 / 5
Completeness	Explicitly answers both 'what' (drive a native macOS app via cua-driver MCP/CLI, snapshot, click/type/scroll, verify) and 'when' ('Use when the user asks you to operate, drive, automate, or perform a GUI task in a real macOS application on the host') with concrete trigger examples.	5 / 5
Trigger Term Quality	Strong natural-language coverage ('operate, drive, automate, perform a GUI task') plus concrete example phrases users would say; just shy of a 5 because it lacks synonym/file-extension style enumeration.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche — backgrounded GUI automation of native macOS apps via cua-driver — with distinct, specific triggers and minimal overlap with other skills.	5 / 5
	Total	19 / 20 Passed

Validation

68%

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 — 11 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (949 lines); consider splitting into references/ and linking	Warning
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 2 missing	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	11 / 16 Passed

Repository: ThinkInAIXYZ/deepchat
Path: plugins/cua/vendor/cua-driver/source/Skills/cua-driver/SKILL.md
Commit: f886d6e

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.