figure-composer

Compose one publication-grade multi-panel figure. Entry from a one-line claim + data files, OR from an existing figure via `derive_outline_prompt` (you read the PNG). Runs a per-figure loop: outline (12-col grid, per-panel ask + label_budget) → render each panel with `panel_task` (loading `figure-style`), one at a time or parallelized → tile + stamp letters with `compose_figure` → adversarial composite self-review with two-tier feedback (Tier-1 outline_revisions / Tier-2 per-panel violations) → regen affected panels, ≤3 rounds. Helpers: panel_task / compose_figure / compose_crops / composite_review_task / derive_outline_prompt. For one standalone plot use `figure-style`; for whole-paper figure ordering use `paper-narrative`.

Quality

86%

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

88%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 high-quality, actionable skill body with a clear sequenced workflow, executable code, and built-in validation/feedback loops. The only soft spots are slight redundancy around single-vs-parallel rendering and that detailed figure-style rules live in an external skill rather than bundled references.

Suggestions

Trim the repeated single-agent-vs-parallel explanation in §2 to a single statement, since the same tradeoff is restated in both paragraphs.

If figure-style §7.1 and §9.2 are load-bearing for this skill, consider bundling those rule excerpts as local reference files so the skill is navigable without loading the sibling skill.

Dimension	Reasoning	Score
Conciseness	The body is largely lean and assumes Claude's competence (executable snippets, direct step headings), with only minor over-explanation such as re-asserting the single-agent/parallel tradeoff across multiple sections that could be tightened.	4 / 5
Actionability	Provides fully executable, copy-paste-ready guidance — `exec(open('figure-composer/kernel.py').read())`, the `panel_task`/`compose_figure` calls with exact arguments, and the PIL crop snippet — covering the common cases concretely.	5 / 5
Workflow Clarity	The pipeline is explicitly sequenced (Setup → outline → render → compose → §3.5 visual QA → §4 adversarial loop → anti-patterns) with validation checkpoints (look-before-review crop pass, review_schema() JSON validation, max-3-round convergence floor) and clear error-recovery feedback loops.	5 / 5
Progressive Disclosure	Well-structured with clear section headers and one-level-deep pointers to the figure-style skill (§7.1, §9.2) rather than inlining its rules, though the actual figure-style/kernel.py references are external skill references not bundled here, so navigation is good but not fully self-contained.	4 / 5
	Total	18 / 20 Passed

Description

85%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, specific description that clearly states the capability, the two entry modes, and the routing to sibling skills. Its main weakness is trigger-term quality: it leans on internal helper names rather than the natural phrases a user would utter.

Suggestions

Add plain-language trigger phrases a user would actually say (e.g. 'Use when you need to make a multi-panel publication figure from data', 'publication figure', 'panel figure') so the description fires on natural requests, not just on the helper jargon.

Consider surfacing 'publication-grade figure', 'multi-panel figure', and 'compose figure from CSV/parquet' as explicit keywords alongside the helper names.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'Compose one publication-grade multi-panel figure', 'render each panel with panel_task', 'tile + stamp letters with compose_figure', 'adversarial composite self-review' — covering the full pipeline comprehensively.	5 / 5
Completeness	Explicitly answers both what ('Compose one publication-grade multi-panel figure') and when to use it ('For one standalone plot use figure-style; for whole-paper figure ordering use paper-narrative'), plus concrete entry-trigger guidance ('Entry from a one-line claim + data files, OR from an existing figure').	5 / 5
Trigger Term Quality	It names domain actions and helper keywords but the natural user-facing trigger phrases are thin — a user is more likely to say 'make a figure' or 'publication figure' than the helper/skill jargon ('derive_outline_prompt', 'panel_task') that dominates the description.	3 / 5
Distinctiveness Conflict Risk	It carves a clear niche — multi-panel figure composition — and explicitly routes the adjacent cases elsewhere (standalone plot → figure-style, whole-paper ordering → paper-narrative), giving it minimal conflict risk.	5 / 5
	Total	18 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: UnicomAI/wanwu
Path: configs/microservice/bff-service/configs/agent-skills/claude-science/figure-composer/SKILL.md
Commit: 030c20b

Reviewed: 18 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.