Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a comprehensive, highly actionable skill that provides everything needed to produce Metis Strategy whitepapers — from brand asset paths to rendering commands to citation verification workflows. The workflow is well-sequenced with strong validation gates. The main weakness is verbosity: the citation audit section and human review guidance could be significantly tightened without losing clarity, and some instructions explain things Claude already knows (basic text editing, what HTML tags are).
Suggestions
Tighten the citation audit section (Step 8) — the hallucination pattern list and Part B instructions could be condensed by ~40% while preserving all actionable guidance.
Remove the 'How to spot-edit whitepaper.html' section (Step 9) — this explains basic text editing concepts to the user, not to Claude, and doesn't belong in a skill file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is quite long (~500+ lines) and includes some information Claude could infer (e.g., explaining what Notepad and VS Code are, basic HTML editing guidance). However, most content is domain-specific brand rules, asset paths, and workflow details that Claude genuinely needs. The citation audit section (Part B) is notably verbose with extensive hallucination pattern lists that could be tightened. | 2 / 3 |
Actionability | Highly actionable throughout — provides exact file paths, complete CSS snippets, executable HTML templates, specific bash commands, concrete config.json schemas, and copy-paste-ready code for metric cards, callout boxes, SVG overlays, and Playwright rendering. Every instruction is specific enough to execute directly. | 3 / 3 |
Workflow Clarity | The 11-step workflow is clearly sequenced with explicit validation checkpoints: outline approval gate (step 3), citation audit with two mandatory parts (step 8), human review before rendering (step 9), and final visual verification (step 11). Feedback loops are present — fix issues in HTML, re-render, and re-verify. The mandatory 'wait for user approval' gates prevent premature drafting. | 3 / 3 |
Progressive Disclosure | Cleanly structured with a main SKILL.md overview that delegates detailed content to three reference files (content-patterns.md, page-layouts.md, diagram-builders.md) and a scripts folder. References are one level deep and clearly signaled at the top. The main file provides enough context to understand the full workflow without needing to read the references first. | 3 / 3 |
Total | 11 / 12 Passed |