メニュー
Defines operational modes for agent work - planning (read-only analysis), editing (full write access), and interactive (guided development). Helps agents self-regulate tool usage based on current phase.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SOC 職業分類に基づく
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。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/majiayu000/claude-skill-registry --skill operational-modesコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
LLM token logprobs and calibration. Per-decision confidence, ECE, Brier, reliability diagrams, low-confidence triage.
Analyze LLM token logprobs and calibration. Use for per-decision confidence, ECE, Brier scores, reliability diagrams, and low-confidence triage.
回顾最近 N 天的 Claude Code 使用记录——扫描原始会话数据，按主题分组汇总"我都做了什么"，并从个人操作系统视角输出模式、风险与增删建议。当用户说 /recap、"看看我这几天做了什么"、"回顾一下我最近的会话"、"这两天我用 claude 干了啥"、"活动回顾" 时使用。
| name | operational-modes |
| description | Defines operational modes for agent work - planning (read-only analysis), editing (full write access), and interactive (guided development). Helps agents self-regulate tool usage based on current phase. |
| version | 1.0.0 |
| model | sonnet |
| invoked_by | agent |
| user_invocable | true |
| tools | ["Read","Glob","Grep"] |
| best_practices | ["Start in planning mode for unfamiliar tasks","Explicitly transition between modes","Respect mode constraints on tool usage","Use interactive mode for user-guided development","Document mode transitions in task logs"] |
| error_handling | graceful |
| streaming | supported |
Operational modes help agents self-regulate their behavior based on the current phase of work. This prevents premature code changes during analysis, ensures focused work, and creates clear phase transitions.
Purpose: Analysis and planning without making changes.
Characteristics:
Allowed Tools:
Read - Read filesGlob - Find filesGrep - Search contentBash - Read-only commands (ls, cat, git status, git log)Excluded Tools:
Write - No file creationEdit - No file modificationBash - No commands that modify stateNotebookEdit - No notebook changesWhen to Use:
Example Prompt:
You are operating in PLANNING MODE.
Your task is to analyze the codebase but NOT write any code.
Focus on:
- Understanding the current implementation
- Identifying affected components
- Creating a comprehensive plan
- Documenting your findings
Do NOT:
- Create new files
- Modify existing files
- Run commands that change state
Purpose: Full implementation capability.
Characteristics:
Allowed Tools:
WriteEdit - Modify filesBash - All safe commands (build, test, lint)NotebookEdit - Modify notebooksExcluded Tools:
When to Use:
Example Prompt:
You are operating in EDITING MODE.
You have full capability to:
- Create new files
- Modify existing files
- Run build/test commands
Guidelines:
- Follow TDD: write test first, then implement
- Follow project conventions
- Run tests after changes
- Keep changes focused on the task
Purpose: User-guided development with frequent checkpoints.
Characteristics:
Allowed Tools:
Behavior:
When to Use:
Example Prompt:
You are operating in INTERACTIVE MODE.
Before any significant action:
1. Explain what you plan to do
2. Present alternatives if they exist
3. Wait for user confirmation
4. Make the change
5. Report the result
Always keep the user informed and in control.
Purpose: Single-task execution without follow-up.
Characteristics:
Allowed Tools:
Behavior:
When to Use:
Example Prompt:
You are operating in ONE-SHOT MODE.
Complete the following task fully in a single execution:
- Analyze requirements
- Implement solution
- Validate/test
- Summarize changes
Do not expect follow-up interaction.
Transition when:
## Mode Transition: Planning -> Editing
**Planning Summary**:
- Analyzed: [files/components]
- Plan: [brief description]
- Affected: [components list]
**Ready to implement**: [yes/no]
**Blockers**: [none / list blockers]
Transitioning to EDITING MODE.
Transition when:
## Mode Transition: Editing -> Planning
**Reason**: Discovered unexpected dependency
**Need to analyze**: [what needs investigation]
Returning to PLANNING MODE for analysis.
Transition when:
## Mode Transition: -> Interactive
**Reason**: Multiple valid approaches identified
**Options**:
1. Approach A - [pros/cons]
2. Approach B - [pros/cons]
Entering INTERACTIVE MODE for user decision.
When using tasks, note the current mode:
TaskCreate({
subject: 'Implement user validation',
description: '...',
metadata: { mode: 'editing' },
});
TaskUpdate({
taskId: '1',
metadata: { mode: 'planning', reason: 'Need architectural review' },
});
Before changing modes, verify:
Entering Planning Mode:
Entering Editing Mode:
Entering Interactive Mode:
Agent Response:
This is a significant task. I'll start in PLANNING MODE.
## Planning Mode Analysis
### Current Understanding
Reading authentication-related files...
- auth.ts - Main authentication logic
- session.ts - Session management
- middleware/auth.ts - Auth middleware
### Observations
- Current: Custom JWT implementation
- Dependencies: 12 files import from auth.ts
- Tests: 24 tests in auth.test.ts
### Plan
1. Identify all auth touchpoints
2. Design new interface (minimal breaking changes)
3. Implement with backwards compatibility
4. Migrate consumers gradually
### Questions
Before proceeding to EDITING MODE:
1. Should I preserve the current API or is breaking change OK?
2. Target deadline/timeline?
Awaiting your input before transitioning modes.
</usage_example>
Before starting:
Read .claude/context/memory/learnings.md
After completing:
.claude/context/memory/learnings.md.claude/context/memory/issues.md.claude/context/memory/decisions.mdASSUME INTERRUPTION: If it's not in memory, it didn't happen.