[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M69AT2AR4ZXMCD19MM2HET3Q":3,"skill-same:01M69AT2AR4ZXMCD19MM2HET3Q":77},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":23,"publishedAt":32,"linkOnly":33,"markdownContent":34,"markdownBody":35,"frontMatter":36,"refName":47,"commitSha":48,"skillDirectory":49,"forksCount":50,"repositoryHtmlUrl":51,"repositoryHomepageUrl":52,"ownerHtmlUrl":53,"ownerBlogUrl":54,"ownerDescription":55,"updatedAt":56,"install":57,"resources":70},"01M69AT2AR4ZXMCD19MM2HET3Q","copilot-cli-modes",null,"Comprehensive guide to all Copilot CLI running modes: Interactive, Plan, Autopilot, Fleet, Research, Chronicle, and Delegate. Use when asked about CLI modes, workflows, or how to use Copilot CLI effectively.","microsoft\u002Fhaste","microsoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6154722?v=4","01M698VYDXZ6JW5WH3CRV650R8","Microsoft","01M69ASXS77EYG722CCRAQWQTQ",110,"MIT","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fcopilot-cli-modes\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000021","资料检索与研究","SCENARIO","使用场景",[24,25,26,27,28,29,30,31],"cli","copilot","copilot cli modes","docs","github","research","workflows","chronicle","2026-11-10 10:12:30",false,"---\nname: copilot-cli-modes\ndescription: \"Comprehensive guide to all Copilot CLI running modes: Interactive, Plan, Autopilot, Fleet, Research, Chronicle, and Delegate. Use when asked about CLI modes, workflows, or how to use Copilot CLI effectively.\"\nsource: \"GitHub Docs — Copilot CLI concepts and best practices\"\ndomain: \"workflows\"\nlevel: \"foundational\"\nagents: [\"backend-dev\", \"gis\", \"ui\", \"orchestrator\"]\ncreated_date: \"2026-04-27\"\nlast_validated: \"\"\nvalidated_by: \"\"\nstatus: \"draft\"\n---\n\n# Copilot CLI Running Modes\n\n## Overview\n\nCopilot CLI provides 7 distinct running modes, each optimized for different tasks. Understanding when to use each mode is the key to effective AI-assisted development. Modes can be combined — for example, Plan → Autopilot + Fleet is a common power workflow.\n\n---\n\n## Mode Quick Reference\n\n| Mode | Trigger | Autonomy | Modifies Code? | Best For |\n|------|---------|----------|----------------|----------|\n| **Interactive** | Default | Low — asks at every step | Yes (with permission) | Day-to-day coding, Q&A, small changes |\n| **Plan** | `Shift+Tab` or `\u002Fplan` | Low — creates plan, waits for approval | No (planning only) | Complex features, multi-file refactors |\n| **Autopilot** | `Shift+Tab` (cycle) or `--autopilot` | High — works until done | Yes (needs `--allow-all`) | Well-defined tasks, CI, batch ops |\n| **Fleet** | `\u002Ffleet` prefix | High — parallel subagents | Yes | Large parallelizable tasks |\n| **Research** | `\u002Fresearch` | High — autonomous research | No (report only) | Deep technical investigation |\n| **Chronicle** | `\u002Fchronicle` | N\u002FA — reads history | No | Standups, tips, self-improvement |\n| **Delegate** | `\u002Fdelegate` | Full — runs in cloud | Yes (creates PR) | Async work, other repos |\n\n---\n\n## 1. Interactive Mode (Default)\n\n### What It Is\nThe standard back-and-forth mode. You prompt, Copilot responds, you guide the next step. Copilot asks for permission before making changes and checks in with you at decision points.\n\n### When To Use\n- Day-to-day coding tasks\n- Quick bug fixes\n- Single-file changes\n- Exploratory questions about the codebase\n- Tasks where you want control at every step\n\n### Usage Scenarios\n\n**Codebase onboarding:**\n```\nHow is logging configured in this project?\nWhat's the pattern for adding a new API endpoint?\nExplain the authentication flow\n```\n\n**Quick fix:**\n```\nThe login form doesn't validate email format. Fix the validation in src\u002Fcomponents\u002FLoginForm.tsx\n```\n\n**Git operations:**\n```\nCreate a PR for this branch with a detailed description\nRebase this branch against main\n```\n\n### Key Commands\n- Type your prompt and press Enter\n- `Shift+Tab` to cycle to other modes\n- `\u002Fmodel` to switch AI models mid-session\n- `\u002Fclear` or `\u002Fnew` to start fresh between unrelated tasks\n\n### Tips\n- Keep sessions focused — use `\u002Fclear` between unrelated tasks\n- Use `\u002Fcontext` to see how much context window you've used\n- Copilot reads `copilot-instructions.md` automatically\n\n---\n\n## 2. Plan Mode\n\n### What It Is\nA structured planning mode where Copilot analyzes your request, asks clarifying questions, and produces a detailed implementation plan with tasks and checkboxes — **before writing any code**. Plans are saved to `plan.md` in your session folder.\n\n### When To Use\n- Complex multi-file features\n- Refactoring with many touch points\n- New feature implementation\n- Any task where you want to review the approach before execution\n\n### When NOT To Use\n- Quick bug fixes\n- Single-file changes\n- Simple questions\n\n### How To Activate\n- Press `Shift+Tab` to toggle into plan mode (prompts show \"plan\" indicator)\n- Or use `\u002Fplan` command from normal mode\n\n### Usage Scenarios\n\n**Feature planning:**\n```\n\u002Fplan Add OAuth2 authentication with Google and GitHub providers\n```\n\n**Refactoring:**\n```\n\u002Fplan Migrate all class components to functional components with hooks\n```\n\n**Architecture change:**\n```\n\u002Fplan Extract the payment processing logic into a separate microservice\n```\n\n### What Happens\n1. Copilot analyzes your request and codebase\n2. Asks clarifying questions to align on requirements\n3. Creates a structured plan with checkboxes\n4. Saves plan to `plan.md`\n5. Waits for your approval before implementing\n6. After approval, you can:\n   - **Implement manually** — `Proceed with the plan`\n   - **Implement on autopilot** — \"Accept plan and build on autopilot\"\n   - **Implement on autopilot + fleet** — \"Accept plan and build on autopilot + \u002Ffleet\"\n\n### Viewing & Editing Plans\n- `Ctrl+Y` — open plan in your default Markdown editor\n- `\u002Fsession plan` — view current plan in CLI\n\n### The Power Workflow: Explore → Plan → Code → Commit\n```\n1. Explore:  \"Read the authentication files but don't write code yet\"\n2. Plan:    \u002Fplan Implement password reset flow\n3. Review:  Check the plan, suggest modifications\n4. Build:   \"Proceed with the plan\"  (or autopilot)\n5. Verify:  \"Run the tests and fix any failures\"\n6. Commit:  \"Commit these changes with a descriptive message\"\n```\n\n---\n\n## 3. Autopilot Mode\n\n### What It Is\nHands-off autonomous mode. Copilot works through a task without waiting for your input after each step, continuing until the task is complete, a problem occurs, or you press `Ctrl+C`.\n\n### When To Use\n- Well-defined tasks with clear goals\n- Implementing a plan you've already reviewed\n- Batch operations (writing tests, fixing lints)\n- CI\u002FCD workflows and scripting\n- Large tasks that require many steps\n\n### When NOT To Use\n- Open-ended exploration\n- Tasks requiring nuanced judgment calls\n- Vague or ambiguous instructions\n- Tasks where you want to guide each step\n\n### How To Activate\n- Press `Shift+Tab` to cycle to autopilot mode\n- Or use `--autopilot` flag from command line\n- Or accept a plan with \"Accept plan and build on autopilot\"\n\n### Usage Scenarios\n\n**Implement a reviewed plan:**\n```\n# After creating a plan in plan mode:\n\"Accept plan and build on autopilot\"\n```\n\n**Programmatic usage (CI\u002Fscripts):**\n```bash\ncopilot --autopilot --yolo --max-autopilot-continues 10 -p \"Write unit tests for all functions in src\u002Futils\u002F\"\n```\n\n**Batch linting:**\n```bash\ncopilot --autopilot --allow-all -p \"Run the linter, fix all errors, and commit the changes\"\n```\n\n### Permissions\nWhen entering autopilot, you're prompted:\n1. **Enable all permissions** (recommended) — equivalent to `--allow-all` \u002F `--yolo`\n2. **Continue with limited permissions** — auto-denies tool requests needing approval\n3. **Cancel**\n\nUse `\u002Fallow-all` or `\u002Fyolo` mid-session to grant full permissions later.\n\n### Safety Controls\n- `--max-autopilot-continues N` — cap the number of autonomous steps (prevents runaway loops)\n- `Ctrl+C` — stop the agent at any time\n- Each continuation shows premium request usage: `Continuing autonomously (3 premium requests)`\n\n### Comparison Table\n| Flag | What It Does | Autonomy Level |\n|------|-------------|----------------|\n| `--allow-all` \u002F `--yolo` | Auto-approves all tool permissions | Still interactive (asks at decision points) |\n| `--no-ask-user` | Suppresses clarifying questions | Semi-autonomous (no extra premium requests) |\n| `--autopilot` | Full autonomous execution | Fully autonomous (continues until done) |\n\n---\n\n## 4. Fleet Mode (`\u002Ffleet`)\n\n### What It Is\nParallel task execution. The main agent acts as orchestrator, breaking a large task into independent subtasks and assigning them to subagents that run simultaneously. Each subagent has its own context window.\n\n### When To Use\n- Large tasks with multiple independent steps\n- Writing tests for many modules at once\n- Refactoring several unrelated files\n- Updating dependencies across packages\n- Any parallelizable work where speed matters\n\n### When NOT To Use\n- Sequential tasks where each step depends on the previous one\n- Small, focused single-step tasks\n- Tasks where you need fine-grained control over order\n\n### How To Activate\nPrefix your prompt with `\u002Ffleet`:\n```\n\u002Ffleet Write comprehensive unit tests for all controllers in src\u002Fcontrollers\u002F\n```\n\n### Usage Scenarios\n\n**Parallel test writing:**\n```\n\u002Ffleet Write unit tests for UserService, OrderService, and PaymentService. Each test file should cover happy paths, error cases, and edge cases.\n```\n\n**Multi-file refactoring:**\n```\n\u002Ffleet Rename all instances of \"userId\" to \"accountId\" across the following directories: src\u002Fmodels\u002F, src\u002Fservices\u002F, src\u002Fcontrollers\u002F\n```\n\n**Documentation generation:**\n```\n\u002Ffleet Generate JSDoc comments for all exported functions in src\u002Futils\u002F, src\u002Fhelpers\u002F, and src\u002Flib\u002F\n```\n\n**Using specific agents per subtask:**\n```\n\u002Ffleet Use @tester to create tests for src\u002Fauth\u002F. Use @technical-writer to update the API docs in docs\u002F. Use @security-auditor to review src\u002Fpayments\u002F.\n```\n\n**Using specific models per subtask:**\n```\n\u002Ffleet Use Claude Opus 4.5 to refactor the database layer. Use GPT-5.3-Codex to generate the migration scripts.\n```\n\n### Combining with Autopilot\nThe most powerful workflow: Plan → Autopilot + Fleet\n1. `Shift+Tab` into plan mode\n2. Create a detailed implementation plan\n3. Select \"Accept plan and build on autopilot + \u002Ffleet\"\n4. Copilot decomposes the plan, runs subtasks in parallel, and works to completion\n\n### Cost Consideration\nFleet spawns multiple subagents, each consuming premium requests independently. A `\u002Ffleet` task may use more total premium requests than the same task done sequentially. Use `\u002Fmodel` to check your current model's multiplier.\n\n---\n\n## 5. Research Mode (`\u002Fresearch`)\n\n### What It Is\nA specialized deep-research agent that produces comprehensive Markdown reports with citations. Unlike normal chat (optimized for quick answers), `\u002Fresearch` is optimized for thoroughness — reports can be hundreds of lines with architecture diagrams, code snippets, and citations.\n\n### When To Use\n- Deep technical investigation\n- Understanding codebase architecture\n- Comparing technologies or approaches\n- Understanding how a library\u002Fframework works internally\n- Cross-repository research\n\n### When NOT To Use\n- Quick questions (\"what does this function do?\")\n- When you need code changes (research produces reports, not edits)\n- Time-sensitive interactions (research takes longer)\n\n### How To Activate\n```\n\u002Fresearch TOPIC\n```\n\n### Usage Scenarios\n\n**Codebase architecture:**\n```\n\u002Fresearch What is the architecture of this codebase?\n```\n→ Produces architecture diagrams, component breakdowns, data flow descriptions\n\n**Technology comparison:**\n```\n\u002Fresearch What's the difference between JWT and session-based authentication? Include pros, cons, and when to use each.\n```\n→ Narrative explanation with trade-off tables\n\n**Internal patterns:**\n```\n\u002Fresearch How are feature flags implemented at our organization?\n```\n→ Searches org repos first, maps internal patterns\n\n**Deep-dive into a component:**\n```\n\u002Fresearch How is the session management system implemented in this repo? Trace the full flow from login to token refresh.\n```\n→ Walks through actual code, follows imports, traces call chains\n\n**How a technology works:**\n```\n\u002Fresearch How does React implement concurrent rendering? Show the actual source code and internal architecture.\n```\n→ Fetches real source from GitHub, prioritizes code over docs\n\n### Viewing & Sharing Reports\n- `Ctrl+Y` — open the latest report in your editor\n- `\u002Fshare gist research` — save as a GitHub gist\n- `\u002Fshare file research [PATH]` — save to a local file\n- Reports stored at: `~\u002F.copilot\u002Fsession-state\u002FSESSION-ID\u002Fresearch\u002F`\n\n### Report Types\nThe agent adapts its output format based on query type:\n| Query Type | Format | Example |\n|-----------|--------|---------|\n| Process\u002Fhow-to | Step-by-step guide | \"How do I add an endpoint?\" |\n| Conceptual | Narrative + trade-offs | \"What's the difference between X and Y?\" |\n| Technical deep-dive | Architecture diagrams + code | \"How is X implemented?\" |\n\n**Tip**: Be explicit about the report type you want. \"Give me a technical deep-dive into X with architecture diagrams\" produces better results than \"What is X?\"\n\n---\n\n## 6. Chronicle Mode (`\u002Fchronicle`)\n\n### What It Is\nA session history analysis tool. Copilot records every session locally (prompts, responses, tools used, files modified) in a SQLite database. Chronicle reads this history to generate standups, tips, and self-improving custom instructions.\n\n### When To Use\n- Start of day: generate standup report\n- Periodically: discover productivity tips\n- When Copilot keeps making the same mistake: generate corrective instructions\n- To recall past work: search your coding history\n\n### How To Activate\n> **Note**: Chronicle is experimental. Enable with `\u002Fexperimental on` or `--experimental`.\n\n### Subcommands\n\n**Standup report:**\n```\n\u002Fchronicle standup\n\u002Fchronicle standup last 3 days\n```\n→ Summarizes recent work: branch names, PR links, status checks\n\n**Personalized tips:**\n```\n\u002Fchronicle tips\n```\n→ Analyzes your usage patterns and suggests features\u002Fworkflows you're underusing\n\n**Self-improving instructions:**\n```\n\u002Fchronicle improve\n```\n→ Analyzes sessions where Copilot misunderstood you or there was lots of back-and-forth, then generates custom instructions to fix the patterns\n\n**Rebuild session index:**\n```\n\u002Fchronicle reindex\n```\n→ Rebuilds the session store from disk (after deletion, migration, or corruption)\n\n### Usage Scenarios\n\n**Monday morning standup:**\n```\n\u002Fchronicle standup last 3 days\n```\n→ \"Last Friday you worked on the auth refactor (branch: feat\u002Foauth2), created PR #42, and fixed 3 test failures. Thursday you debugged the payment timeout issue...\"\n\n**Discovering unused features:**\n```\n\u002Fchronicle tips\n```\n→ \"You rarely use plan mode before complex tasks. Try \u002Fplan for multi-file changes — models achieve higher success rates with a concrete plan.\"\n\n**Fixing recurring miscommunication:**\n```\n\u002Fchronicle improve\n```\n→ Generates a custom instruction like: \"When I say 'fix the tests', I mean fix failing tests only — do not refactor passing tests or add new ones.\"\n\n### Data Location\n```\n~\u002F.copilot\u002Fsession-state\u002F{session-id}\u002F\n├── events.jsonl          # Full session history\n├── workspace.yaml        # Metadata\n├── plan.md               # Plans (if created)\n└── research\u002F             # Research reports\n\n~\u002F.copilot\u002Fsession-store.db  # SQLite index for chronicle\n```\n\n### Session Management\n- `\u002Fsession` — view current session info\n- `\u002Fsession checkpoints` — view compaction checkpoints\n- `\u002Fsession plan` — view current plan\n- `copilot --continue` — resume last session\n- `copilot --resume` — pick a session to resume\n\n---\n\n## 7. Delegate Mode (`\u002Fdelegate`)\n\n### What It Is\nOffloads a task to Copilot cloud agent, which runs asynchronously and creates a pull request with the results. You can continue working locally while the cloud agent works in the background.\n\n### When To Use\n- Tangential tasks you don't want to block on\n- Documentation updates\n- Refactoring separate modules\n- Changes to other repositories\n- Long-running tasks you can review later\n\n### When NOT To Use\n- Core feature work requiring tight control\n- Active debugging\n- Interactive exploration\n\n### How To Activate\n```\n\u002Fdelegate PROMPT\n```\n\n### Usage Scenarios\n\n**Tangential task:**\n```\n\u002Fdelegate Add dark mode support to the settings page\n```\n→ Cloud agent creates a PR while you keep coding locally\n\n**Documentation:**\n```\n\u002Fdelegate Update the API documentation to reflect the new authentication endpoints\n```\n\n**Cross-repo work:**\n```\n\u002Fdelegate In the frontend repo, update the API client to use the new \u002Fv2\u002Fusers endpoint\n```\n\n---\n\n## Mode Composition Cheat Sheet\n\n| Scenario | Mode Combination | Command |\n|----------|-----------------|---------|\n| Quick fix | Interactive | Just type the prompt |\n| Complex feature | Plan → Autopilot | `Shift+Tab` to plan, then \"Accept plan and build on autopilot\" |\n| Large parallelizable task | Plan → Autopilot + Fleet | Plan, then \"Accept plan and build on autopilot + \u002Ffleet\" |\n| Deep investigation | Research | `\u002Fresearch How does X work?` |\n| CI automation | Autopilot (programmatic) | `copilot --autopilot --yolo --max-autopilot-continues 10 -p \"...\"` |\n| Background work | Delegate | `\u002Fdelegate Update the docs` |\n| Morning ritual | Chronicle | `\u002Fchronicle standup last 3 days` |\n| Productivity review | Chronicle | `\u002Fchronicle tips` then `\u002Fchronicle improve` |\n\n## Decision Tree\n\n```\nIs the task well-defined?\n├── No  → Is it a question?\n│        ├── Quick question → Interactive mode\n│        └── Deep investigation → \u002Fresearch\n└── Yes → Is it complex (multi-file)?\n          ├── No  → Interactive mode\n          └── Yes → Plan mode first\n                    └── Plan approved?\n                        ├── Parallelizable? → Autopilot + \u002Ffleet\n                        ├── Sequential?     → Autopilot\n                        └── Tangential?     → \u002Fdelegate\n```\n","# Copilot CLI Running Modes\n\n## Overview\n\nCopilot CLI provides 7 distinct running modes, each optimized for different tasks. Understanding when to use each mode is the key to effective AI-assisted development. Modes can be combined — for example, Plan → Autopilot + Fleet is a common power workflow.\n\n---\n\n## Mode Quick Reference\n\n| Mode | Trigger | Autonomy | Modifies Code? | Best For |\n|------|---------|----------|----------------|----------|\n| **Interactive** | Default | Low — asks at every step | Yes (with permission) | Day-to-day coding, Q&A, small changes |\n| **Plan** | `Shift+Tab` or `\u002Fplan` | Low — creates plan, waits for approval | No (planning only) | Complex features, multi-file refactors |\n| **Autopilot** | `Shift+Tab` (cycle) or `--autopilot` | High — works until done | Yes (needs `--allow-all`) | Well-defined tasks, CI, batch ops |\n| **Fleet** | `\u002Ffleet` prefix | High — parallel subagents | Yes | Large parallelizable tasks |\n| **Research** | `\u002Fresearch` | High — autonomous research | No (report only) | Deep technical investigation |\n| **Chronicle** | `\u002Fchronicle` | N\u002FA — reads history | No | Standups, tips, self-improvement |\n| **Delegate** | `\u002Fdelegate` | Full — runs in cloud | Yes (creates PR) | Async work, other repos |\n\n---\n\n## 1. Interactive Mode (Default)\n\n### What It Is\nThe standard back-and-forth mode. You prompt, Copilot responds, you guide the next step. Copilot asks for permission before making changes and checks in with you at decision points.\n\n### When To Use\n- Day-to-day coding tasks\n- Quick bug fixes\n- Single-file changes\n- Exploratory questions about the codebase\n- Tasks where you want control at every step\n\n### Usage Scenarios\n\n**Codebase onboarding:**\n```\nHow is logging configured in this project?\nWhat's the pattern for adding a new API endpoint?\nExplain the authentication flow\n```\n\n**Quick fix:**\n```\nThe login form doesn't validate email format. Fix the validation in src\u002Fcomponents\u002FLoginForm.tsx\n```\n\n**Git operations:**\n```\nCreate a PR for this branch with a detailed description\nRebase this branch against main\n```\n\n### Key Commands\n- Type your prompt and press Enter\n- `Shift+Tab` to cycle to other modes\n- `\u002Fmodel` to switch AI models mid-session\n- `\u002Fclear` or `\u002Fnew` to start fresh between unrelated tasks\n\n### Tips\n- Keep sessions focused — use `\u002Fclear` between unrelated tasks\n- Use `\u002Fcontext` to see how much context window you've used\n- Copilot reads `copilot-instructions.md` automatically\n\n---\n\n## 2. Plan Mode\n\n### What It Is\nA structured planning mode where Copilot analyzes your request, asks clarifying questions, and produces a detailed implementation plan with tasks and checkboxes — **before writing any code**. Plans are saved to `plan.md` in your session folder.\n\n### When To Use\n- Complex multi-file features\n- Refactoring with many touch points\n- New feature implementation\n- Any task where you want to review the approach before execution\n\n### When NOT To Use\n- Quick bug fixes\n- Single-file changes\n- Simple questions\n\n### How To Activate\n- Press `Shift+Tab` to toggle into plan mode (prompts show \"plan\" indicator)\n- Or use `\u002Fplan` command from normal mode\n\n### Usage Scenarios\n\n**Feature planning:**\n```\n\u002Fplan Add OAuth2 authentication with Google and GitHub providers\n```\n\n**Refactoring:**\n```\n\u002Fplan Migrate all class components to functional components with hooks\n```\n\n**Architecture change:**\n```\n\u002Fplan Extract the payment processing logic into a separate microservice\n```\n\n### What Happens\n1. Copilot analyzes your request and codebase\n2. Asks clarifying questions to align on requirements\n3. Creates a structured plan with checkboxes\n4. Saves plan to `plan.md`\n5. Waits for your approval before implementing\n6. After approval, you can:\n   - **Implement manually** — `Proceed with the plan`\n   - **Implement on autopilot** — \"Accept plan and build on autopilot\"\n   - **Implement on autopilot + fleet** — \"Accept plan and build on autopilot + \u002Ffleet\"\n\n### Viewing & Editing Plans\n- `Ctrl+Y` — open plan in your default Markdown editor\n- `\u002Fsession plan` — view current plan in CLI\n\n### The Power Workflow: Explore → Plan → Code → Commit\n```\n1. Explore:  \"Read the authentication files but don't write code yet\"\n2. Plan:    \u002Fplan Implement password reset flow\n3. Review:  Check the plan, suggest modifications\n4. Build:   \"Proceed with the plan\"  (or autopilot)\n5. Verify:  \"Run the tests and fix any failures\"\n6. Commit:  \"Commit these changes with a descriptive message\"\n```\n\n---\n\n## 3. Autopilot Mode\n\n### What It Is\nHands-off autonomous mode. Copilot works through a task without waiting for your input after each step, continuing until the task is complete, a problem occurs, or you press `Ctrl+C`.\n\n### When To Use\n- Well-defined tasks with clear goals\n- Implementing a plan you've already reviewed\n- Batch operations (writing tests, fixing lints)\n- CI\u002FCD workflows and scripting\n- Large tasks that require many steps\n\n### When NOT To Use\n- Open-ended exploration\n- Tasks requiring nuanced judgment calls\n- Vague or ambiguous instructions\n- Tasks where you want to guide each step\n\n### How To Activate\n- Press `Shift+Tab` to cycle to autopilot mode\n- Or use `--autopilot` flag from command line\n- Or accept a plan with \"Accept plan and build on autopilot\"\n\n### Usage Scenarios\n\n**Implement a reviewed plan:**\n```\n# After creating a plan in plan mode:\n\"Accept plan and build on autopilot\"\n```\n\n**Programmatic usage (CI\u002Fscripts):**\n```bash\ncopilot --autopilot --yolo --max-autopilot-continues 10 -p \"Write unit tests for all functions in src\u002Futils\u002F\"\n```\n\n**Batch linting:**\n```bash\ncopilot --autopilot --allow-all -p \"Run the linter, fix all errors, and commit the changes\"\n```\n\n### Permissions\nWhen entering autopilot, you're prompted:\n1. **Enable all permissions** (recommended) — equivalent to `--allow-all` \u002F `--yolo`\n2. **Continue with limited permissions** — auto-denies tool requests needing approval\n3. **Cancel**\n\nUse `\u002Fallow-all` or `\u002Fyolo` mid-session to grant full permissions later.\n\n### Safety Controls\n- `--max-autopilot-continues N` — cap the number of autonomous steps (prevents runaway loops)\n- `Ctrl+C` — stop the agent at any time\n- Each continuation shows premium request usage: `Continuing autonomously (3 premium requests)`\n\n### Comparison Table\n| Flag | What It Does | Autonomy Level |\n|------|-------------|----------------|\n| `--allow-all` \u002F `--yolo` | Auto-approves all tool permissions | Still interactive (asks at decision points) |\n| `--no-ask-user` | Suppresses clarifying questions | Semi-autonomous (no extra premium requests) |\n| `--autopilot` | Full autonomous execution | Fully autonomous (continues until done) |\n\n---\n\n## 4. Fleet Mode (`\u002Ffleet`)\n\n### What It Is\nParallel task execution. The main agent acts as orchestrator, breaking a large task into independent subtasks and assigning them to subagents that run simultaneously. Each subagent has its own context window.\n\n### When To Use\n- Large tasks with multiple independent steps\n- Writing tests for many modules at once\n- Refactoring several unrelated files\n- Updating dependencies across packages\n- Any parallelizable work where speed matters\n\n### When NOT To Use\n- Sequential tasks where each step depends on the previous one\n- Small, focused single-step tasks\n- Tasks where you need fine-grained control over order\n\n### How To Activate\nPrefix your prompt with `\u002Ffleet`:\n```\n\u002Ffleet Write comprehensive unit tests for all controllers in src\u002Fcontrollers\u002F\n```\n\n### Usage Scenarios\n\n**Parallel test writing:**\n```\n\u002Ffleet Write unit tests for UserService, OrderService, and PaymentService. Each test file should cover happy paths, error cases, and edge cases.\n```\n\n**Multi-file refactoring:**\n```\n\u002Ffleet Rename all instances of \"userId\" to \"accountId\" across the following directories: src\u002Fmodels\u002F, src\u002Fservices\u002F, src\u002Fcontrollers\u002F\n```\n\n**Documentation generation:**\n```\n\u002Ffleet Generate JSDoc comments for all exported functions in src\u002Futils\u002F, src\u002Fhelpers\u002F, and src\u002Flib\u002F\n```\n\n**Using specific agents per subtask:**\n```\n\u002Ffleet Use @tester to create tests for src\u002Fauth\u002F. Use @technical-writer to update the API docs in docs\u002F. Use @security-auditor to review src\u002Fpayments\u002F.\n```\n\n**Using specific models per subtask:**\n```\n\u002Ffleet Use Claude Opus 4.5 to refactor the database layer. Use GPT-5.3-Codex to generate the migration scripts.\n```\n\n### Combining with Autopilot\nThe most powerful workflow: Plan → Autopilot + Fleet\n1. `Shift+Tab` into plan mode\n2. Create a detailed implementation plan\n3. Select \"Accept plan and build on autopilot + \u002Ffleet\"\n4. Copilot decomposes the plan, runs subtasks in parallel, and works to completion\n\n### Cost Consideration\nFleet spawns multiple subagents, each consuming premium requests independently. A `\u002Ffleet` task may use more total premium requests than the same task done sequentially. Use `\u002Fmodel` to check your current model's multiplier.\n\n---\n\n## 5. Research Mode (`\u002Fresearch`)\n\n### What It Is\nA specialized deep-research agent that produces comprehensive Markdown reports with citations. Unlike normal chat (optimized for quick answers), `\u002Fresearch` is optimized for thoroughness — reports can be hundreds of lines with architecture diagrams, code snippets, and citations.\n\n### When To Use\n- Deep technical investigation\n- Understanding codebase architecture\n- Comparing technologies or approaches\n- Understanding how a library\u002Fframework works internally\n- Cross-repository research\n\n### When NOT To Use\n- Quick questions (\"what does this function do?\")\n- When you need code changes (research produces reports, not edits)\n- Time-sensitive interactions (research takes longer)\n\n### How To Activate\n```\n\u002Fresearch TOPIC\n```\n\n### Usage Scenarios\n\n**Codebase architecture:**\n```\n\u002Fresearch What is the architecture of this codebase?\n```\n→ Produces architecture diagrams, component breakdowns, data flow descriptions\n\n**Technology comparison:**\n```\n\u002Fresearch What's the difference between JWT and session-based authentication? Include pros, cons, and when to use each.\n```\n→ Narrative explanation with trade-off tables\n\n**Internal patterns:**\n```\n\u002Fresearch How are feature flags implemented at our organization?\n```\n→ Searches org repos first, maps internal patterns\n\n**Deep-dive into a component:**\n```\n\u002Fresearch How is the session management system implemented in this repo? Trace the full flow from login to token refresh.\n```\n→ Walks through actual code, follows imports, traces call chains\n\n**How a technology works:**\n```\n\u002Fresearch How does React implement concurrent rendering? Show the actual source code and internal architecture.\n```\n→ Fetches real source from GitHub, prioritizes code over docs\n\n### Viewing & Sharing Reports\n- `Ctrl+Y` — open the latest report in your editor\n- `\u002Fshare gist research` — save as a GitHub gist\n- `\u002Fshare file research [PATH]` — save to a local file\n- Reports stored at: `~\u002F.copilot\u002Fsession-state\u002FSESSION-ID\u002Fresearch\u002F`\n\n### Report Types\nThe agent adapts its output format based on query type:\n| Query Type | Format | Example |\n|-----------|--------|---------|\n| Process\u002Fhow-to | Step-by-step guide | \"How do I add an endpoint?\" |\n| Conceptual | Narrative + trade-offs | \"What's the difference between X and Y?\" |\n| Technical deep-dive | Architecture diagrams + code | \"How is X implemented?\" |\n\n**Tip**: Be explicit about the report type you want. \"Give me a technical deep-dive into X with architecture diagrams\" produces better results than \"What is X?\"\n\n---\n\n## 6. Chronicle Mode (`\u002Fchronicle`)\n\n### What It Is\nA session history analysis tool. Copilot records every session locally (prompts, responses, tools used, files modified) in a SQLite database. Chronicle reads this history to generate standups, tips, and self-improving custom instructions.\n\n### When To Use\n- Start of day: generate standup report\n- Periodically: discover productivity tips\n- When Copilot keeps making the same mistake: generate corrective instructions\n- To recall past work: search your coding history\n\n### How To Activate\n> **Note**: Chronicle is experimental. Enable with `\u002Fexperimental on` or `--experimental`.\n\n### Subcommands\n\n**Standup report:**\n```\n\u002Fchronicle standup\n\u002Fchronicle standup last 3 days\n```\n→ Summarizes recent work: branch names, PR links, status checks\n\n**Personalized tips:**\n```\n\u002Fchronicle tips\n```\n→ Analyzes your usage patterns and suggests features\u002Fworkflows you're underusing\n\n**Self-improving instructions:**\n```\n\u002Fchronicle improve\n```\n→ Analyzes sessions where Copilot misunderstood you or there was lots of back-and-forth, then generates custom instructions to fix the patterns\n\n**Rebuild session index:**\n```\n\u002Fchronicle reindex\n```\n→ Rebuilds the session store from disk (after deletion, migration, or corruption)\n\n### Usage Scenarios\n\n**Monday morning standup:**\n```\n\u002Fchronicle standup last 3 days\n```\n→ \"Last Friday you worked on the auth refactor (branch: feat\u002Foauth2), created PR #42, and fixed 3 test failures. Thursday you debugged the payment timeout issue...\"\n\n**Discovering unused features:**\n```\n\u002Fchronicle tips\n```\n→ \"You rarely use plan mode before complex tasks. Try \u002Fplan for multi-file changes — models achieve higher success rates with a concrete plan.\"\n\n**Fixing recurring miscommunication:**\n```\n\u002Fchronicle improve\n```\n→ Generates a custom instruction like: \"When I say 'fix the tests', I mean fix failing tests only — do not refactor passing tests or add new ones.\"\n\n### Data Location\n```\n~\u002F.copilot\u002Fsession-state\u002F{session-id}\u002F\n├── events.jsonl          # Full session history\n├── workspace.yaml        # Metadata\n├── plan.md               # Plans (if created)\n└── research\u002F             # Research reports\n\n~\u002F.copilot\u002Fsession-store.db  # SQLite index for chronicle\n```\n\n### Session Management\n- `\u002Fsession` — view current session info\n- `\u002Fsession checkpoints` — view compaction checkpoints\n- `\u002Fsession plan` — view current plan\n- `copilot --continue` — resume last session\n- `copilot --resume` — pick a session to resume\n\n---\n\n## 7. Delegate Mode (`\u002Fdelegate`)\n\n### What It Is\nOffloads a task to Copilot cloud agent, which runs asynchronously and creates a pull request with the results. You can continue working locally while the cloud agent works in the background.\n\n### When To Use\n- Tangential tasks you don't want to block on\n- Documentation updates\n- Refactoring separate modules\n- Changes to other repositories\n- Long-running tasks you can review later\n\n### When NOT To Use\n- Core feature work requiring tight control\n- Active debugging\n- Interactive exploration\n\n### How To Activate\n```\n\u002Fdelegate PROMPT\n```\n\n### Usage Scenarios\n\n**Tangential task:**\n```\n\u002Fdelegate Add dark mode support to the settings page\n```\n→ Cloud agent creates a PR while you keep coding locally\n\n**Documentation:**\n```\n\u002Fdelegate Update the API documentation to reflect the new authentication endpoints\n```\n\n**Cross-repo work:**\n```\n\u002Fdelegate In the frontend repo, update the API client to use the new \u002Fv2\u002Fusers endpoint\n```\n\n---\n\n## Mode Composition Cheat Sheet\n\n| Scenario | Mode Combination | Command |\n|----------|-----------------|---------|\n| Quick fix | Interactive | Just type the prompt |\n| Complex feature | Plan → Autopilot | `Shift+Tab` to plan, then \"Accept plan and build on autopilot\" |\n| Large parallelizable task | Plan → Autopilot + Fleet | Plan, then \"Accept plan and build on autopilot + \u002Ffleet\" |\n| Deep investigation | Research | `\u002Fresearch How does X work?` |\n| CI automation | Autopilot (programmatic) | `copilot --autopilot --yolo --max-autopilot-continues 10 -p \"...\"` |\n| Background work | Delegate | `\u002Fdelegate Update the docs` |\n| Morning ritual | Chronicle | `\u002Fchronicle standup last 3 days` |\n| Productivity review | Chronicle | `\u002Fchronicle tips` then `\u002Fchronicle improve` |\n\n## Decision Tree\n\n```\nIs the task well-defined?\n├── No  → Is it a question?\n│        ├── Quick question → Interactive mode\n│        └── Deep investigation → \u002Fresearch\n└── Yes → Is it complex (multi-file)?\n          ├── No  → Interactive mode\n          └── Yes → Plan mode first\n                    └── Plan approved?\n                        ├── Parallelizable? → Autopilot + \u002Ffleet\n                        ├── Sequential?     → Autopilot\n                        └── Tangential?     → \u002Fdelegate\n```\n",{"name":5,"level":37,"agents":38,"domain":30,"source":43,"status":44,"description":7,"created_date":45,"validated_by":46,"last_validated":46},"foundational",[39,40,41,42],"backend-dev","gis","ui","orchestrator","GitHub Docs — Copilot CLI concepts and best practices","draft","2026-04-27","","main","20513f6bc8d842697a2192d068aa48ef3c22c957",".github\u002Fskills\u002Fcopilot-cli-modes",19,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste","https:\u002F\u002Fmicrosoft.github.io\u002Fhaste\u002F","https:\u002F\u002Fgithub.com\u002Fmicrosoft","https:\u002F\u002Fopensource.microsoft.com","Open source projects and samples from Microsoft","2026-11-10 14:49:19",{"source":8,"skillName":5,"commands":58,"prompt":68,"zipAvailable":69},[59,62,65],{"tool":60,"command":61},"npx","npx skills add microsoft\u002Fhaste --skill copilot-cli-modes",{"tool":63,"command":64},"bunx","bunx skills add microsoft\u002Fhaste --skill copilot-cli-modes",{"tool":66,"command":67},"pnpm","pnpm dlx skills add microsoft\u002Fhaste --skill copilot-cli-modes","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fcopilot-cli-modes\u002FSKILL.md",true,[71],{"path":72,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":75,"sourceHtmlUrl":76},"references\u002Ffabric-embed.md","REFERENCE","参考文档",5052,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FDataHub-App-for-Microsoft-Copilot-Agents-Cowork-and-Autopilots\u002Fblob\u002Fmain\u002Fapp\u002F.agents\u002Fskills\u002Fapp-validation\u002Freferences\u002Ffabric-embed.md",[78,92,102,114,122,131],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":32},"01M69ATF7EZ2CBCPFKQ99NZJEM","validation-diagnostics","Validation and diagnostic skill for HASTE. Compare planned vs implemented work, generate diagnostic reports, and feed misses back into skill refinement. Use when: 'validate implementation', 'compare to spec', 'diagnostic report', 'drift analysis', 'coverage check', 'implementation review'.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fvalidation-diagnostics\u002FSKILL.md",[],[85,86,87,88,89,90,91],"validation diagnostics","diagnostics","review","spec","security","to spec","validate",{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":32},"01M69ATDJ0SQ0RMHK9148MKYHB","security-analysis","Dependabot and security analysis skill for HASTE. Parse alerts, group related vulnerabilities, apply severity rules, and produce structured reports. Use when: 'Dependabot alert', 'security scan', 'vulnerability triage', 'dependency audit', 'npm audit', 'pip audit', 'CVE analysis'.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fsecurity-analysis\u002FSKILL.md",[],[89,99,100,101],"security analysis","vulnerabilities","triage",{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":32},"01M69ATBBN86366SYW86NM19PT","release-notes","Generate release notes from commit history. Use when asked to draft release notes, create a changelog, or summarize changes for a release.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Frelease-notes\u002FSKILL.md",[],[109,110,111,112,113],"release notes","notes","summarize","commit","documentation",{"skillNo":115,"name":116,"nameZh":6,"description":117,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":118,"categories":119,"keywords":120,"publishedAt":32},"01M69AT9PND3YVN90VZFP0PW2W","imagery-provider-adaptation","Imagery provider adaptation skill for HASTE. Encapsulates provider-specific logic for satellite imagery sources (Planet, Vantor, Airbus, etc.). Use when: 'new imagery provider', 'add source type', 'satellite provider', 'Planet', 'Vantor', 'Airbus', 'Pleiades', 'WorldView', 'SkySat', 'imagery ingestion', 'provider adapter'.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fimagery-provider-adaptation\u002FSKILL.md",[],[121,113],"imagery provider adaptation",{"skillNo":123,"name":124,"nameZh":6,"description":125,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":126,"categories":127,"keywords":128,"publishedAt":32},"01M69AT83XYH0W0A22KJNQSZG3","deterministic-scripts","Deterministic script execution skill for HASTE. Execute scripts instead of free-form LLM behavior for consistent, repeatable operations. Use when: 'run preprocessing', 'convert format', 'parse metadata', 'generate tiles', 'build wheel', 'deploy functions'. Prevents hallucinated scripts.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fdeterministic-scripts\u002FSKILL.md",[],[129,130],"deterministic scripts","deploy",{"skillNo":132,"name":133,"nameZh":6,"description":134,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":135,"categories":136,"keywords":137,"publishedAt":32},"01M69AT6GZFSRAHX1169T5FZNZ","dependency-update","Guide for safely updating project dependencies. Use when asked to update packages, upgrade dependencies, or check for outdated libraries.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhaste\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fdependency-update\u002FSKILL.md",[],[138,89],"dependency update"]