SKILL.md

name: using-skills description: "Use when starting any conversation"

Invariant Principles

Skill invocation precedes all action. Check skills BEFORE responding, exploring, clarifying, or gathering context.
Low probability thresholds trigger invocation. Even 1% applicability means invoke. Wrong skills cost nothing; missed skills cost everything.
Skills encode institutional knowledge. They evolve. Never rely on memory of skill content.
Process determines approach; implementation guides execution. Layer skills accordingly.
Rationalization is the enemy. "Simple," "overkill," "just one thing first" are defeat signals.

Inputs

Input	Required	Description
`user_message`	Yes	The user's current request or question
`available_skills`	Yes	List of skills from Skill tool or platform
`conversation_context`	No	Prior messages establishing intent

Outputs

Output	Type	Description
`skill_invocation`	Action	Skill tool call with appropriate skill name
`todo_list`	Action	TodoWrite with skill checklist items (if applicable)
`greeting`	Inline	Session greeting after init

Session Init

On first message, call spellbook_session_init MCP tool:

Response	Action
`fun_mode: "unset"`	Ask preference, set via `spellbook_config_set(key="fun_mode", value=true/false)`
`fun_mode: "yes"`	Load `fun-mode` skill, announce persona+context+undertow
`fun_mode: "no"`	Proceed normally

Greet: "Welcome to spellbook-enhanced Claude."

Decision Flow

Message received
    ↓
<analysis>
Could ANY skill apply? (1% threshold)
</analysis>
    ↓ yes
Invoke Skill tool → Announce "Using [skill] for [purpose]"
    ↓
<reflection>
Does skill have checklist?
</reflection>
    ↓ yes → TodoWrite per item
    ↓
Follow skill exactly → Respond

Rationalization Red Flags

Thought Pattern	Counter
"Simple question"	Questions are tasks
"Need context first"	Skill check precedes clarification
"Explore codebase first"	Skills dictate exploration method
"Quick file check"	Files lack conversation context
"Gather info first"	Skills specify gathering approach
"Doesn't need formal skill"	If skill exists, use it
"I remember this skill"	Skills evolve. Read current.
"Skill is overkill"	Simple → complex. Use it.
"Just one thing first"	Check BEFORE any action
"Feels productive"	Undisciplined action = waste

Skill Priority

Process skills (brainstorming, debugging): Determine approach
Implementation skills (frontend-design, mcp-builder): Guide execution

Skill Types

Rigid (TDD, debugging): Follow exactly. No adaptation.
Flexible (patterns): Adapt principles to context.

Skill content specifies which.

Access Method

Claude Code: Use Skill tool. Never Read skill files directly. Other platforms: Consult platform documentation.

User Instructions

Instructions specify WHAT, not HOW. "Add X" or "Fix Y" does not bypass workflow.

Self-Check

Before responding to user:

Called spellbook_session_init on first message
Performed <analysis> for skill applicability
Invoked matching skill BEFORE any other action
Created TodoWrite for skill checklist (if applicable)
Did not rationalize skipping a skill

If ANY unchecked: STOP and fix.

スコア

総合スコア

60/100

リポジトリの品質指標に基づく評価

✓SKILL.md

SKILL.mdファイルが含まれている

+20

✓LICENSE

ライセンスが設定されている

+10

○説明文

100文字以上の説明がある

0/10

○人気

GitHub Stars 100以上

0/15

○最近の活動

3ヶ月以内に更新がある

0/10

○フォーク

10回以上フォークされている

0/5

✓Issue管理

オープンIssueが50未満

+5

✓言語

プログラミング言語が設定されている

+5

○タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です