canvas-designer

Core canvas design skill covering project management, multimedia principles, AI image generation, web image search, and design marker processing. Load for any canvas design task. CRITICAL - When user message contains [@design_canvas_project:...] or [@design_marker:...] mentions, or when the user wants to generate video/animation/clip on a canvas project, you MUST load this skill first before any operations.

Quality

79%

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 ./backend/super-magic/agents/skills/canvas-designer/SKILL.md

Quality

Content

71%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.

A highly actionable skill with excellent executable examples and mostly clear workflows including a strong dry-run feedback loop. Its weaknesses are a verbose inline Prompt Engineering essay that over-explains general concepts Claude already knows, and several inline references to a reference/ bundle whose files are not present.

Suggestions

Move the general Prompt Engineering essay (specificity-vs-adjectives, sensory stacking, scene-type table, intent-to-prompt steps) into a separate reference file (e.g. reference/image/prompt-engineering.md) and keep only canvas-specific prompt rules inline, to cut token bloat and raise conciseness.

Create the referenced bundle files (reference/image/image-search.md, reference/image/design-marker.md, reference/video/generation.md, reference/video/parameters-and-errors.md) or remove the dead links, so the progressive-disclosure navigation actually resolves.

Add an explicit post-generation verification step for batch/overwrite paths (e.g. after generate_canvas_images with element_id retries, confirm via result.content that all tasks succeeded before proceeding) to lift workflow_clarity to score 5.

Dimension	Reasoning	Score
Conciseness	The Prompt Engineering section (lines 174-354) is a lengthy essay on general prompt craft — creative-brief framing, specificity vs adjectives, sensory stacking, scene-type tables — much of which a capable model already knows and which competes with the canvas-specific context window. The skill is mostly efficient elsewhere (parameter tables, decision tree), but this substantial padding pulls it toward the 'mostly efficient but includes unnecessary explanation' anchor.	3 / 5
Actionability	Provides fully executable, copy-paste-ready code for every major path (text-to-image with a complete multi-task example, image-to-image, multi-reference style transfer, retry-by-element_id, anchor-then-expand multi-view) plus concrete parameter tables and a decision tree, with specific examples covering the common cases.	5 / 5
Workflow Clarity	Sequences are clear and most have validation checkpoints: restore_canvas_media has an explicit dry-run-then-confirm feedback loop, failed tasks return element_id for in-place retry, and the multi-view anchor-then-expand pattern is a well-sequenced two-call workflow. It does not quite reach score 5 because several destructive/batch paths (e.g. overwriting via element_id, multi-task generation) lack an explicit post-generation verification step beyond reading result.content.	4 / 5
Progressive Disclosure	The body is well-sectioned and does signal one-level-deep references (reference/image/image-search.md, reference/image/design-marker.md, reference/video/generation.md, reference/video/parameters-and-errors.md), but no reference/ directory or any of those files actually exist in the bundle, so the navigation points to missing materials. The Prompt Engineering section is also a large body of content that reads as though it belongs in a dedicated reference file rather than inline in SKILL.md.	3 / 5
	Total	15 / 20 Passed

Description

87%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.

A strong description that clearly states both capabilities and load triggers, anchored by distinctive mention syntaxes that make conflict risk very low. The main weakness is that capabilities are framed as domains rather than concrete named actions, and a few natural user keywords (poster, illustration) are absent.

Suggestions

Reframe capability domains as concrete actions where possible (e.g. 'generate AI images on a canvas', 'search and place web images', 'process design markers') to match the specificity anchor at score 5.

Add a few more natural trigger keywords users say beyond the mention syntax (e.g. 'poster', 'illustration', 'cover image') to broaden trigger-term coverage.

Keep the 'CRITICAL ... MUST load this skill first' trigger clause — it is the strongest part and should be preserved.

Dimension	Reasoning	Score
Specificity	Lists several specific capability areas ('project management, multimedia principles, AI image generation, web image search, and design marker processing') rather than vague abstractions, but these are domains/categories rather than concrete named actions like 'fill forms' or 'extract text', leaving minor gaps in coverage.	4 / 5
Completeness	Explicitly answers both what ('Core canvas design skill covering project management, multimedia principles, AI image generation, web image search, and design marker processing') and when with concrete trigger phrases ('Load for any canvas design task', 'When user message contains [@design_canvas_project:...] or [@design_marker:...] mentions, or when the user wants to generate video/animation/clip').	5 / 5
Trigger Term Quality	Includes natural trigger phrases users would say ('generate video/animation/clip', 'canvas design task', 'web image search') plus the exact mention syntaxes '[@design_canvas_project:...]' and '[@design_marker:...]', but is missing a few common synonyms (e.g. 'poster', 'illustration', 'image') a user might naturally say.	4 / 5
Distinctiveness Conflict Risk	The distinctive mention syntaxes '[@design_canvas_project:...]' and '[@design_marker:...]' create a clear niche with minimal conflict risk, and the explicit 'you MUST load this skill first' boundary on those mentions sharply separates it from generic image/design skills.	5 / 5
	Total	18 / 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 (671 lines); consider splitting into references/ and linking	Warning
relative_links	Relative link issues: 5 missing, 5 deeper-than-1-level	Warning

	Total	14 / 16 Passed

Repository: dtyq/magic
Path: backend/super-magic/agents/skills/canvas-designer/SKILL.md
Commit: f9973c5

Reviewed: 1 month 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.