Add requirement-clarifier: three-mode triage station

[yingzhengzhang06-sys]

What this skill does

A three-mode triage station for ambiguous requests. It does not write code, write articles, or make decisions for you. It turns fuzzy intent into an executable spec, a task pile into a routed execution order, and an existing plan into an improvement checklist. Then it points to the next skill.

This skill is a router that dispatches to three sub-skills:

• growme-mode — turn vague intent into a requirement spec
• change-mode — turn a task pile into a routed execution plan
• improve-mode — turn an existing plan into an improvement checklist

Why it should be in skills.sh

• Native Chinese scenarios — cross-runtime neutral (Claude Code / Codex / OpenCode / OpenClaw / Hermes)
• Three-mode suite — three independent sub-skills + router; most existing skills are single-mode (grill-me, brainstorming, interview-me, office-hours are all single-mode)
• 30+ skill routing table as a public asset — installing this skill is installing a 'full skill routing map'
• Structured output templates for spec / triage / review (3 templates)
• 8 test-prompts.json live samples + 3 anti-pattern examples
• Luban 9-dimension rubric score: 90/100 (3 independent-agent live replays + 1 cross-mode serial test)

What makes it different from similar skills

Dimension	Similar (mattpocock / obra / addyosmani)	This skill
Mode	Single clarification (grill-me / brainstorming / interview-me / office-hours)	Three-mode unified
Routing	Direct to plan / write code	Routes to specific execution skills (30+ skill routing table)
Language	English-first	Native Chinese scenarios, cross-runtime neutral
Form	Single SKILL.md	Suite: 1 router + 3 independent sub-skills
Output	Free-form text	Structured output templates
Verification	Mostly missing	8 test-prompts + 3 anti-pattern examples

Group placement

I added a new Methodology group in skills.sh.json since this skill doesn't fit React / Vercel / Design. Happy to move it to a different group or rename if maintainers prefer.

Checklist

• SKILL.md with valid frontmatter (name, description, license, metadata)
• AGENTS.md (124 lines, agent-readable compiled output)
• metadata.json (version, organization, date, abstract, references)
• README.md (Chinese, house-style)
• README.en.md (English version)
• LICENSE (MIT)
• 3 sub-modes with own SKILL.md + references/
• 3 anti-pattern examples
• test-prompts.json (8 P1-P5 samples)
• Cross-runtime compatible
• No private paths, no API keys, no fake data
• Source repo: https://github.com/yingzhengzhang06-sys/requirement-clarifier

Known issues

• Showcase GIF pending: vhs 0.11 + ttyd 1.7.7 compatibility fault on macOS only records first frame. references/vhs-fault-note.md documents the issue. Skill functionality unaffected.
• Live user replay (not agent simulation) pending v3.0.1

[vercel]

Someone is attempting to deploy a commit to the Vercel Team on Vercel.

A member of the Team first needs to authorize it.