ワンクリックでManusで任意のスキルを実行

new-skill

スター2

フォーク0

更新日2026年6月13日 07:44

Use when adding a new skill to the spine repo — scaffolds a conventions-compliant SKILL.md, wires it into plugin.json, the README, and the skill index, and runs the validator until green. Meta-tooling that dogfoods spine's own contribution rules so extensions land correctly the first time.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

AndrewTtofi

AndrewTtofi/spine

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

new-skill — scaffold a spine skill, wired and valid

Adding a skill means more than writing a SKILL.md — it must be wired into the manifest, the README, and the index, or scripts/validate.mjs fails. This skill does the whole job to the repo's own quality bar. Unlike the lifecycle skills, it operates on the spine repo itself, not a user's .spine/.

Steps

Read the conventions. Load CLAUDE.md (the wiring rules the validator enforces) and CONTRIBUTING.md, and skim an existing SKILL.md as the template for voice and structure.
Interview. Settle, one question at a time: the skill's name (which must equal its folder), the phase/problem it owns, explicit trigger phrases for the description, whether it's a lifecycle skill (and thus declares Spine I/O), and what sibling reference files — if any — it needs.
Write skills/<name>/SKILL.md. Frontmatter name matching the folder and a description with "Use when …" triggers. Keep the body thin: a one-line intent, ## Steps, ## Spine I/O (for lifecycle skills), and ## Notes. Push depth into sibling files and link to them — progressive disclosure.
Wire it in (all three, or the validator fails):
- add ./skills/<name> to .claude-plugin/plugin.json
- link it in the README's Skills section
- list it in skills/README.md
Validate. Run node scripts/validate.mjs and fix until it prints All N skills valid.
Hand off. Suggest committing, and point at ship to open the PR.

What it touches

Unlike a lifecycle skill, this reads and writes the spine repo's own files — CLAUDE.md, CONTRIBUTING.md, .claude-plugin/plugin.json, README.md, skills/README.md, and the new skills/<name>/ directory. It does not touch a user's .spine/.

Notes

The validator is the test. A skill isn't done until scripts/validate.mjs is green — frontmatter name must equal the folder, and all three wirings present.
Match the existing skills' voice: senior-engineer terse, no ceremony.
Don't bloat SKILL.md. If it's getting long, a sibling reference file is the answer.

このリポジトリの他の Skills

同じリポジトリ

ship

AndrewTtofi/spine

Use when verified work is ready to land — turns it into a clean branch, commits, and a pull request whose body is drawn from the Spine (criteria, decisions, evidence). Refuses to ship work that hasn't passed verify. Reads the Spine; writes the shipped status back to it.

2026-06-142

raise-debrief

AndrewTtofi/spine

Use right after a VC meeting to turn raw memory into a structured debrief — a signal read (interest vs. hesitation vs. pushback), the open diligence queue with draft answers, a send-ready follow-up email, and a clear recommended next step. Run within 24 hours; the follow-up email degrades fast. Use when asked to "debrief the meeting", "how did the pitch go", "write the follow-up email", or "what do I do after the meeting with <fund>".

2026-06-142

raise-report

AndrewTtofi/spine

Use to see the whole raise at a glance and to produce an investor-ready dossier — scans the .spine/raise/ namespace, builds a pipeline status tracker (every fund's scores, stage, and next action), and assembles a consolidated dossier suitable for sharing with an investor or advisor. Use when asked to "show the pipeline", "raise status", "track the funds", "what's stale", or "generate the investor report".

2026-06-142

raise-term

AndrewTtofi/spine

Use to analyse a VC term sheet — benchmarks every clause against current market norms, flags founder-unfriendly terms, scores founder-friendliness out of 100, and produces a ranked negotiation priority list with suggested pushback language. Informational only, not legal advice. Use when asked to "analyse this term sheet", "review the term sheet", "is this term sheet fair", or "what should I negotiate".

2026-06-142

raise-outreach

AndrewTtofi/spine

Use to turn a fund dossier into ready-to-send outreach — a cold email, a warm-intro request your contact can forward, and a LinkedIn DM, each built on the fund's own thesis language, plus a fund-specific one-pager. Refuses to draft for a no-go fund. Drafts only — it never sends. Use when asked to "draft outreach for <fund>", "write the intro email", "cold email this VC", or "make a one-pager for <fund>".

2026-06-142

raise-pitch

AndrewTtofi/spine

Use to prepare for pitching — full-spectrum prep in one skill. With no fund, it runs the stress-test: ten lethal questions a hostile VC would ask (know what to defend) and ten strongest signals a champion would champion (know what to lead with). With a fund, it builds a timed meeting plan with segment scripts, preemptive answers to that fund's diligence questions, and red-flag rebuttals. Use when asked to "prep me to pitch", "stress-test my pitch", "destroy my startup", "what are my strongest signals", or "prep for the meeting with <fund>".

2026-06-142

name	new-skill
description	Use when adding a new skill to the spine repo — scaffolds a conventions-compliant SKILL.md, wires it into plugin.json, the README, and the skill index, and runs the validator until green. Meta-tooling that dogfoods spine's own contribution rules so extensions land correctly the first time.

new-skill — scaffold a spine skill, wired and valid

Steps

Read the conventions. Load CLAUDE.md (the wiring rules the validator enforces) and CONTRIBUTING.md, and skim an existing SKILL.md as the template for voice and structure.
Interview. Settle, one question at a time: the skill's name (which must equal its folder), the phase/problem it owns, explicit trigger phrases for the description, whether it's a lifecycle skill (and thus declares Spine I/O), and what sibling reference files — if any — it needs.
Write skills/<name>/SKILL.md. Frontmatter name matching the folder and a description with "Use when …" triggers. Keep the body thin: a one-line intent, ## Steps, ## Spine I/O (for lifecycle skills), and ## Notes. Push depth into sibling files and link to them — progressive disclosure.
Wire it in (all three, or the validator fails):
- add ./skills/<name> to .claude-plugin/plugin.json
- link it in the README's Skills section
- list it in skills/README.md
Validate. Run node scripts/validate.mjs and fix until it prints All N skills valid.
Hand off. Suggest committing, and point at ship to open the PR.

What it touches

Notes

The validator is the test. A skill isn't done until scripts/validate.mjs is green — frontmatter name must equal the folder, and all three wirings present.
Match the existing skills' voice: senior-engineer terse, no ceremony.
Don't bloat SKILL.md. If it's getting long, a sibling reference file is the answer.