Skill Build Kit — QC Checklist

Run this checklist after building both assets. Every item is pass/fail. Fix failures before shipping.

1. YAML Frontmatter

• [ ] name field present (lowercase, hyphenated)
• [ ] description field is multi-line, includes what it does + when to use it
• [ ] description includes triggers list with natural language variations
• [ ] metadata block present with author, version, date
• [ ] Version follows semantic format (e.g., 3.0.0)

2. Core Principle

• [ ] Named and bolded — not a vague philosophy
• [ ] Specific operating rule, not a mission statement
• [ ] 2-3 sentence explanation of what happens if the rule isn't followed

3. What This Skill Does

• [ ] Intro paragraph states what user pastes and what they get back
• [ ] Three jobs defined, each with a bold name and practical description
• [ ] Jobs describe output, not process ("produces a structured brief" not "analyzes your emails")

4. Section-by-Section Spec

• [ ] Every output section has its own H3 heading
• [ ] Each section defines: what it contains, why it matters, format rules
• [ ] Format rules are specific (table structure, column names, line counts — not "keep it brief")
• [ ] Edge cases addressed (e.g., "If only one email provided, say so — don't guess")

5. Rules

• [ ] Stated as imperatives with bold labels
• [ ] Cover output format constraints (no paragraphs, scannable, tables)
• [ ] Cover scope constraints (one client per run, recency window)
• [ ] Cover quality constraints (don't assume, cite evidence, numbers over labels)
• [ ] Cover edge cases (insufficient data, missing context)
• [ ] Cover naming/file conventions for output

6. Output Format Template

• [ ] Complete markdown template with placeholders
• [ ] Every section header present
• [ ] Every table structure shown with column names
• [ ] Every label/value pair shown
• [ ] Template is copy-pasteable — someone could fill in the blanks

7. Cross-References (Series Skills)

• [ ] Other skills referenced by full name — no abbreviations
• [ ] User reminded the referenced skill exists ("if you don't have it, ask for it")
• [ ] If skill accepts another skill's output as input: mapping section present
• [ ] Mapping section uses table format linking other skill's sections to this skill's detection patterns
• [ ] Both input paths (raw input AND other skill's output) documented as equal

8. What Makes This Different

• [ ] Speaks to the user (not instructions for Claude)
• [ ] Contrasts with the generic alternative ("Most X tools give you Y. This gives you Z.")
• [ ] Uses specific language, not feature-speak
• [ ] 2-3 paragraphs, no more

9. Testing

• [ ] Skill tested on real data (not fabricated input)
• [ ] At least 3 inputs from one real scenario
• [ ] Output matches the section-by-section spec
• [ ] Output format matches the template
• [ ] Edge cases tested (minimal input, missing context)
• [ ] Output is scannable — no section requires more than 5 seconds to absorb

10. Quick-Start Guide

• [ ] Opens with "You just got a free Claude skill. Here's how to use it."
• [ ] Prerequisites listed (Claude Pro, input description)
• [ ] Step 1: Enable Code Execution — exact UI path
• [ ] Step 2: Upload the Skill — exact button names
• [ ] "Running the Skill" section with trigger phrase
• [ ] Tips section — 3-5 tips from real testing, not generic advice
• [ ] "What to Paste" section — accepted input types as a list
• [ ] Footer: Built by Kathryn Brown — Advisory OS

11. Voice (Both Assets)

• [ ] Skill file: direct, rules-oriented, format-explicit, no hedging
• [ ] Quick-start: casual, step-numbered, assumption-free
• [ ] Uses locked terms (skill, run, paste, install) — not alternatives
• [ ] No forbidden constructions (check 02-skill-build-terminology.md)
• [ ] "What Makes This Different" uses brand voice — pattern-revealing, specific, grounded