[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6J8SH186V7VM7STWPAEE7WD":3,"skill-same:01M6J8SH186V7VM7STWPAEE7WD":81},{"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":45,"linkOnly":46,"markdownContent":47,"markdownBody":48,"frontMatter":49,"refName":60,"commitSha":61,"skillDirectory":62,"forksCount":54,"repositoryHtmlUrl":63,"repositoryHomepageUrl":6,"ownerHtmlUrl":64,"ownerBlogUrl":65,"ownerDescription":66,"updatedAt":67,"install":68,"resources":80},"01M6J8SH186V7VM7STWPAEE7WD","chat-plugin-Claude Code Orchestration",null,"Skill for orchestrating Claude Code sessions from OpenClaw. Covers launching, monitoring, multi-turn interaction, lifecycle management, notifications, and parallel work patterns.","Flissel\u002Fvibemind-os","Flissel","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F104565236?v=4","01M6J8Q88NWAJGKPC5TEG9YCNV","Felix Baumann","01M6J8Q8D7XAFY44P70XTJFZBW",1,"MIT","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fskills\u002Fchat-plugin\u002FClaude Code Orchestration\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44],"orchestration","plugin","code","chat plugin Claude Code Orchestration","chat","2026","openclaw","source","for","true","interaction","imported","patterns","repo","work","parallel","lifecycle","claude code sessions","notifications","monitoring","sessions","2026-11-10 10:12:30",false,"---\nname: chat-plugin-Claude Code Orchestration\ndescription: Skill for orchestrating Claude Code sessions from OpenClaw. Covers launching,\n  monitoring, multi-turn interaction, lifecycle management, notifications, and parallel\n  work patterns.\napp: chat-plugin\nrequires_approval: true\nagents:\n- '*'\ninputs: []\nexpected_state:\n  description: Imported skill — behavior matches source repo\n  verification_tool: none\nconfidence: 1.0\nattempts: 0\nsuccesses: 0\nlast_adjusted: '2026-05-11T21:59:41+00:00'\n---\n\n# Claude Code Orchestration\n\nYou orchestrate Claude Code sessions via the `openclaw-claude-code-plugin`. Each session is an autonomous agent that executes code tasks in the background.\n\n---\n\n## 1. Launching sessions\n\n### Mandatory rules\n\n- **Notifications are routed automatically** via `agentChannels` config. Do NOT pass `channel` manually — it bypasses automatic routing.\n- **Always pass `multi_turn: true`** unless the task is a guaranteed one-shot with no possible follow-up.\n- **Name the sessions** with `name` in kebab-case, short and descriptive.\n- **Set `workdir`** to the target project directory, not the agent's workspace.\n\n### Essential parameters\n\n| Parameter | When to use |\n|---|---|\n| `prompt` | Always. Clear and complete instruction. |\n| `name` | Always. Descriptive kebab-case (`fix-auth-bug`, `add-dark-mode`). |\n| `channel` | **Do NOT pass.** Resolved automatically via `agentChannels`. |\n| `workdir` | Always when the project is not in the `defaultWorkdir`. |\n| `multi_turn` | `true` by default unless explicitly one-shot. |\n| `max_budget_usd` | Adjust accordingly: `1-2` for a small fix, `5` for a feature, `10+` for a major refactoring. |\n| `model` | When you want to force a specific model (`\"sonnet\"`, `\"opus\"`). |\n| `system_prompt` | To inject project-specific context. |\n| `permission_mode` | `\"bypassPermissions\"` by default. `\"acceptEdits\"` for more control. |\n\n### Examples\n\n```\n# Simple task\nclaude_launch(\n  prompt: \"Fix the null pointer in src\u002Fauth.ts line 42\",\n  name: \"fix-null-auth\",\n  workdir: \"\u002Fhome\u002Fuser\u002Fprojects\u002Fmyapp\",\n  multi_turn: true,\n  max_budget_usd: 2\n)\n\n# Full feature\nclaude_launch(\n  prompt: \"Implement dark mode toggle in the settings page. Use the existing theme context in src\u002Fcontext\u002Ftheme.tsx. Add a toggle switch component and persist the preference in localStorage.\",\n  name: \"add-dark-mode\",\n  workdir: \"\u002Fhome\u002Fuser\u002Fprojects\u002Fmyapp\",\n  multi_turn: true,\n  max_budget_usd: 5\n)\n\n# Major refactoring\nclaude_launch(\n  prompt: \"Refactor the database layer to use the repository pattern. Migrate all direct Prisma calls in src\u002Fservices\u002F to use repositories in src\u002Frepositories\u002F.\",\n  name: \"refactor-db-repositories\",\n  workdir: \"\u002Fhome\u002Fuser\u002Fprojects\u002Fmyapp\",\n  multi_turn: true,\n  max_budget_usd: 10,\n  model: \"opus\"\n)\n```\n\n### Resume and fork\n\n```\n# Resume a completed session\nclaude_launch(\n  prompt: \"Continue. Also add error handling for the edge cases we discussed.\",\n  resume_session_id: \"fix-null-auth\",\n  multi_turn: true\n)\n\n# Fork to try an alternative approach\nclaude_launch(\n  prompt: \"Try a completely different approach: use middleware instead of decorators.\",\n  resume_session_id: \"refactor-db-repositories\",\n  fork_session: true,\n  name: \"refactor-db-middleware-approach\",\n  multi_turn: true\n)\n```\n\n---\n\n## 2. Monitoring sessions\n\n### List sessions\n\n```\n# All sessions\nclaude_sessions()\n\n# Only running sessions\nclaude_sessions(status: \"running\")\n\n# Completed sessions (for resume)\nclaude_sessions(status: \"completed\")\n```\n\n### View output\n\n```\n# Summary (last 50 lines)\nclaude_output(session: \"fix-null-auth\")\n\n# Full output (up to 200 blocks)\nclaude_output(session: \"fix-null-auth\", full: true)\n\n# Specific last N lines\nclaude_output(session: \"fix-null-auth\", lines: 100)\n```\n\n### Real-time streaming\n\n```\n# Switch to foreground (displays catchup of missed outputs + live stream)\nclaude_fg(session: \"fix-null-auth\")\n\n# Switch back to background (stops the stream, session continues)\nclaude_bg(session: \"fix-null-auth\")\n\n# Detach all foreground sessions from a channel\nclaude_bg()\n```\n\n**Note:** `claude_fg` first displays a catchup \"Catchup (N missed outputs):\" with everything that happened in the background, then starts live streaming.\n\n---\n\n## 3. Multi-turn interaction\n\n### Send a follow-up\n\n```\n# Reply to a Claude question\nclaude_respond(session: \"add-dark-mode\", message: \"Yes, use CSS variables for the theme colors.\")\n\n# Redirect a running session (interrupts the current turn)\nclaude_respond(session: \"add-dark-mode\", message: \"Stop. Use Tailwind dark: classes instead of CSS variables.\", interrupt: true)\n```\n\n### When to auto-respond vs forward to the user\n\n**Auto-respond immediately with `claude_respond`:**\n- Permission requests to read\u002Fwrite files or run bash commands -> `\"Yes, proceed.\"`\n- Confirmations like \"Should I continue?\" -> `\"Yes, continue.\"`\n- Questions about the approach when only one is reasonable -> Respond with the obvious choice\n- Clarification requests about the codebase -> Respond if you know, otherwise `\"Use your best judgment.\"`\n\n**Forward to the user:**\n- Architecture decisions (Redis vs PostgreSQL, REST vs GraphQL...)\n- Destructive operations (deleting files, dropping tables...)\n- Ambiguous requirements not covered by the initial prompt\n- Scope or budget changes (\"This will require refactoring 15 files\")\n- Anything involving credentials, secrets, or production environments\n- When in doubt -> always forward to the user\n\n### Interaction cycle\n\n1. Session launches -> runs in background\n2. Wake event `openclaw system event` arrives when the session is waiting for input\n3. Read the question with `claude_output(session, full: true)`\n4. Decide: auto-respond or forward\n5. If auto-respond: `claude_respond(session, answer)`\n6. If forward: relay the question to the user, wait for their response, then `claude_respond`\n\n---\n\n## 4. Lifecycle management\n\n### Stop a session\n\n```\nclaude_kill(session: \"fix-null-auth\")\n```\n\nUse when:\n- The session is stuck or looping\n- Budget is being wasted on the wrong approach\n- The user requests a stop\n\n### Timeouts\n\n- Idle multi-turn sessions are automatically killed after `idleTimeoutMinutes` (default: 30 min)\n- Completed sessions are garbage-collected after 1h but remain resumable via persisted IDs\n\n### Check the result after completion\n\nWhen a session completes (completion wake event):\n\n1. `claude_output(session: \"xxx\", full: true)` to read the result\n2. Summarize the result to the user\n3. If failed, analyze the error and decide: relaunch, fork, or escalate\n\n---\n\n## 5. Notifications\n\n### Routing\n\nNotifications are routed automatically based on the session's `workdir` using the `agentChannels` plugin config. Each workspace directory maps to a specific channel (e.g., `telegram:seo-bot:123456789`). See [Agent Channels](..\u002F..\u002Fdocs\u002FAGENT_CHANNELS.md) for setup.\n\n### Events\n\n| Event | What happens |\n|---|---|\n| Session starts | Silent |\n| Session in foreground | Real-time stream (text debounced 500ms + tool indicators) |\n| Session completed | Notification to the originating channel |\n| Session failed | Error notification to the originating channel |\n| Waiting for input | `openclaw system event` to wake the orchestrator + `\"Claude asks:\"` in the channel |\n| Response sent | `\"Responded:\"` echo in the channel |\n| Session > 10 min | One-shot reminder |\n| Session limit reached | Notification to the channel |\n\n### Background visibility\n\nEven without foreground, the user sees in their chat:\n- **`Claude asks:`** when Claude asks a question\n- **`Responded:`** when a follow-up is sent\n\nThis makes the conversation transparent without needing foregrounding.\n\n---\n\n## 6. Best practices\n\n### Launch checklist\n\n1. `agentChannels` is configured for this workdir -> notifications arrive\n2. `multi_turn: true` -> interaction is possible after launch\n3. `name` is descriptive -> easy to identify in `claude_sessions`\n4. `workdir` points to the correct project -> Claude Code works in the right directory\n5. `max_budget_usd` is calibrated -> no waste\n\n### Parallel tasks\n\n```\n# Launch multiple sessions on independent tasks\nclaude_launch(prompt: \"Build the frontend auth page\", name: \"frontend-auth\", workdir: \"\u002Fapp\u002Ffrontend\", multi_turn: true, max_budget_usd: 5)\nclaude_launch(prompt: \"Build the backend auth API\", name: \"backend-auth\", workdir: \"\u002Fapp\u002Fbackend\", multi_turn: true, max_budget_usd: 5)\n```\n\n- Respect the `maxSessions` limit (default: 5)\n- Each session must have a unique `name`\n- Monitor each session individually via wake events\n- Don't launch too many sessions in parallel: each session consumes resources and budget\n\n### Error handling\n\n- If a session fails, read the full output to understand why\n- Relaunch with a corrected prompt if the error is in the instruction\n- Fork (`fork_session: true`) if you want to try a different approach without losing context\n- Escalate to the user if the error is outside your control (permissions, network, missing dependencies)\n\n### Reporting results\n\nWhen a session completes:\n1. Read the full result with `claude_output(session, full: true)`\n2. Summarize the changes made to the user\n3. Mention the files modified\u002Fcreated\n4. Flag any issues encountered or remaining TODOs\n\n---\n\n## 7. Anti-patterns\n\n| Anti-pattern | Consequence | Fix |\n|---|---|---|\n| Passing `channel` explicitly | Bypasses automatic routing, notifications may go to wrong bot | Let `agentChannels` handle routing automatically |\n| Forgetting `multi_turn: true` | Unable to send follow-ups with `claude_respond` | Enable `multi_turn` except for explicit one-shots |\n| Not checking the result of a completed session | The user doesn't know what happened | Always read `claude_output` and summarize |\n| Launching too many sessions in parallel | `maxSessions` limit reached, sessions rejected, diluted attention | Respect the limit, prioritize, sequence if necessary |\n| Using the agent's `workdir` instead of the project's | Claude Code works in the wrong directory | Always point to the target project directory |\n| Not naming sessions | Hard to identify them in `claude_sessions` | Always use `name` in kebab-case |\n| Auto-responding to critical questions | Decisions made without the user's approval | When in doubt, forward to the user |\n| Ignoring wake events | Sessions stuck waiting indefinitely | Handle each wake event promptly |\n| Not adjusting the budget | Waste (too high) or interruption (too low) | Calibrate `max_budget_usd` based on complexity |\n\n---\n\n## 8. Quick tool reference\n\n| Tool | Usage | Key parameters |\n|---|---|---|\n| `claude_launch` | Launch a session | `prompt`, `name`, `workdir`, `multi_turn`, `max_budget_usd` |\n| `claude_sessions` | List sessions | `status` (all\u002Frunning\u002Fcompleted\u002Ffailed) |\n| `claude_output` | Read the output | `session`, `full`, `lines` |\n| `claude_fg` | Foreground + live stream | `session` |\n| `claude_bg` | Switch back to background | `session` |\n| `claude_kill` | Kill a session | `session` |\n| `claude_respond` | Send a follow-up | `session`, `message`, `interrupt` |\n| `claude_stats` | Usage metrics | none |\n","# Claude Code Orchestration\n\nYou orchestrate Claude Code sessions via the `openclaw-claude-code-plugin`. Each session is an autonomous agent that executes code tasks in the background.\n\n---\n\n## 1. Launching sessions\n\n### Mandatory rules\n\n- **Notifications are routed automatically** via `agentChannels` config. Do NOT pass `channel` manually — it bypasses automatic routing.\n- **Always pass `multi_turn: true`** unless the task is a guaranteed one-shot with no possible follow-up.\n- **Name the sessions** with `name` in kebab-case, short and descriptive.\n- **Set `workdir`** to the target project directory, not the agent's workspace.\n\n### Essential parameters\n\n| Parameter | When to use |\n|---|---|\n| `prompt` | Always. Clear and complete instruction. |\n| `name` | Always. Descriptive kebab-case (`fix-auth-bug`, `add-dark-mode`). |\n| `channel` | **Do NOT pass.** Resolved automatically via `agentChannels`. |\n| `workdir` | Always when the project is not in the `defaultWorkdir`. |\n| `multi_turn` | `true` by default unless explicitly one-shot. |\n| `max_budget_usd` | Adjust accordingly: `1-2` for a small fix, `5` for a feature, `10+` for a major refactoring. |\n| `model` | When you want to force a specific model (`\"sonnet\"`, `\"opus\"`). |\n| `system_prompt` | To inject project-specific context. |\n| `permission_mode` | `\"bypassPermissions\"` by default. `\"acceptEdits\"` for more control. |\n\n### Examples\n\n```\n# Simple task\nclaude_launch(\n  prompt: \"Fix the null pointer in src\u002Fauth.ts line 42\",\n  name: \"fix-null-auth\",\n  workdir: \"\u002Fhome\u002Fuser\u002Fprojects\u002Fmyapp\",\n  multi_turn: true,\n  max_budget_usd: 2\n)\n\n# Full feature\nclaude_launch(\n  prompt: \"Implement dark mode toggle in the settings page. Use the existing theme context in src\u002Fcontext\u002Ftheme.tsx. Add a toggle switch component and persist the preference in localStorage.\",\n  name: \"add-dark-mode\",\n  workdir: \"\u002Fhome\u002Fuser\u002Fprojects\u002Fmyapp\",\n  multi_turn: true,\n  max_budget_usd: 5\n)\n\n# Major refactoring\nclaude_launch(\n  prompt: \"Refactor the database layer to use the repository pattern. Migrate all direct Prisma calls in src\u002Fservices\u002F to use repositories in src\u002Frepositories\u002F.\",\n  name: \"refactor-db-repositories\",\n  workdir: \"\u002Fhome\u002Fuser\u002Fprojects\u002Fmyapp\",\n  multi_turn: true,\n  max_budget_usd: 10,\n  model: \"opus\"\n)\n```\n\n### Resume and fork\n\n```\n# Resume a completed session\nclaude_launch(\n  prompt: \"Continue. Also add error handling for the edge cases we discussed.\",\n  resume_session_id: \"fix-null-auth\",\n  multi_turn: true\n)\n\n# Fork to try an alternative approach\nclaude_launch(\n  prompt: \"Try a completely different approach: use middleware instead of decorators.\",\n  resume_session_id: \"refactor-db-repositories\",\n  fork_session: true,\n  name: \"refactor-db-middleware-approach\",\n  multi_turn: true\n)\n```\n\n---\n\n## 2. Monitoring sessions\n\n### List sessions\n\n```\n# All sessions\nclaude_sessions()\n\n# Only running sessions\nclaude_sessions(status: \"running\")\n\n# Completed sessions (for resume)\nclaude_sessions(status: \"completed\")\n```\n\n### View output\n\n```\n# Summary (last 50 lines)\nclaude_output(session: \"fix-null-auth\")\n\n# Full output (up to 200 blocks)\nclaude_output(session: \"fix-null-auth\", full: true)\n\n# Specific last N lines\nclaude_output(session: \"fix-null-auth\", lines: 100)\n```\n\n### Real-time streaming\n\n```\n# Switch to foreground (displays catchup of missed outputs + live stream)\nclaude_fg(session: \"fix-null-auth\")\n\n# Switch back to background (stops the stream, session continues)\nclaude_bg(session: \"fix-null-auth\")\n\n# Detach all foreground sessions from a channel\nclaude_bg()\n```\n\n**Note:** `claude_fg` first displays a catchup \"Catchup (N missed outputs):\" with everything that happened in the background, then starts live streaming.\n\n---\n\n## 3. Multi-turn interaction\n\n### Send a follow-up\n\n```\n# Reply to a Claude question\nclaude_respond(session: \"add-dark-mode\", message: \"Yes, use CSS variables for the theme colors.\")\n\n# Redirect a running session (interrupts the current turn)\nclaude_respond(session: \"add-dark-mode\", message: \"Stop. Use Tailwind dark: classes instead of CSS variables.\", interrupt: true)\n```\n\n### When to auto-respond vs forward to the user\n\n**Auto-respond immediately with `claude_respond`:**\n- Permission requests to read\u002Fwrite files or run bash commands -> `\"Yes, proceed.\"`\n- Confirmations like \"Should I continue?\" -> `\"Yes, continue.\"`\n- Questions about the approach when only one is reasonable -> Respond with the obvious choice\n- Clarification requests about the codebase -> Respond if you know, otherwise `\"Use your best judgment.\"`\n\n**Forward to the user:**\n- Architecture decisions (Redis vs PostgreSQL, REST vs GraphQL...)\n- Destructive operations (deleting files, dropping tables...)\n- Ambiguous requirements not covered by the initial prompt\n- Scope or budget changes (\"This will require refactoring 15 files\")\n- Anything involving credentials, secrets, or production environments\n- When in doubt -> always forward to the user\n\n### Interaction cycle\n\n1. Session launches -> runs in background\n2. Wake event `openclaw system event` arrives when the session is waiting for input\n3. Read the question with `claude_output(session, full: true)`\n4. Decide: auto-respond or forward\n5. If auto-respond: `claude_respond(session, answer)`\n6. If forward: relay the question to the user, wait for their response, then `claude_respond`\n\n---\n\n## 4. Lifecycle management\n\n### Stop a session\n\n```\nclaude_kill(session: \"fix-null-auth\")\n```\n\nUse when:\n- The session is stuck or looping\n- Budget is being wasted on the wrong approach\n- The user requests a stop\n\n### Timeouts\n\n- Idle multi-turn sessions are automatically killed after `idleTimeoutMinutes` (default: 30 min)\n- Completed sessions are garbage-collected after 1h but remain resumable via persisted IDs\n\n### Check the result after completion\n\nWhen a session completes (completion wake event):\n\n1. `claude_output(session: \"xxx\", full: true)` to read the result\n2. Summarize the result to the user\n3. If failed, analyze the error and decide: relaunch, fork, or escalate\n\n---\n\n## 5. Notifications\n\n### Routing\n\nNotifications are routed automatically based on the session's `workdir` using the `agentChannels` plugin config. Each workspace directory maps to a specific channel (e.g., `telegram:seo-bot:123456789`). See [Agent Channels](..\u002F..\u002Fdocs\u002FAGENT_CHANNELS.md) for setup.\n\n### Events\n\n| Event | What happens |\n|---|---|\n| Session starts | Silent |\n| Session in foreground | Real-time stream (text debounced 500ms + tool indicators) |\n| Session completed | Notification to the originating channel |\n| Session failed | Error notification to the originating channel |\n| Waiting for input | `openclaw system event` to wake the orchestrator + `\"Claude asks:\"` in the channel |\n| Response sent | `\"Responded:\"` echo in the channel |\n| Session > 10 min | One-shot reminder |\n| Session limit reached | Notification to the channel |\n\n### Background visibility\n\nEven without foreground, the user sees in their chat:\n- **`Claude asks:`** when Claude asks a question\n- **`Responded:`** when a follow-up is sent\n\nThis makes the conversation transparent without needing foregrounding.\n\n---\n\n## 6. Best practices\n\n### Launch checklist\n\n1. `agentChannels` is configured for this workdir -> notifications arrive\n2. `multi_turn: true` -> interaction is possible after launch\n3. `name` is descriptive -> easy to identify in `claude_sessions`\n4. `workdir` points to the correct project -> Claude Code works in the right directory\n5. `max_budget_usd` is calibrated -> no waste\n\n### Parallel tasks\n\n```\n# Launch multiple sessions on independent tasks\nclaude_launch(prompt: \"Build the frontend auth page\", name: \"frontend-auth\", workdir: \"\u002Fapp\u002Ffrontend\", multi_turn: true, max_budget_usd: 5)\nclaude_launch(prompt: \"Build the backend auth API\", name: \"backend-auth\", workdir: \"\u002Fapp\u002Fbackend\", multi_turn: true, max_budget_usd: 5)\n```\n\n- Respect the `maxSessions` limit (default: 5)\n- Each session must have a unique `name`\n- Monitor each session individually via wake events\n- Don't launch too many sessions in parallel: each session consumes resources and budget\n\n### Error handling\n\n- If a session fails, read the full output to understand why\n- Relaunch with a corrected prompt if the error is in the instruction\n- Fork (`fork_session: true`) if you want to try a different approach without losing context\n- Escalate to the user if the error is outside your control (permissions, network, missing dependencies)\n\n### Reporting results\n\nWhen a session completes:\n1. Read the full result with `claude_output(session, full: true)`\n2. Summarize the changes made to the user\n3. Mention the files modified\u002Fcreated\n4. Flag any issues encountered or remaining TODOs\n\n---\n\n## 7. Anti-patterns\n\n| Anti-pattern | Consequence | Fix |\n|---|---|---|\n| Passing `channel` explicitly | Bypasses automatic routing, notifications may go to wrong bot | Let `agentChannels` handle routing automatically |\n| Forgetting `multi_turn: true` | Unable to send follow-ups with `claude_respond` | Enable `multi_turn` except for explicit one-shots |\n| Not checking the result of a completed session | The user doesn't know what happened | Always read `claude_output` and summarize |\n| Launching too many sessions in parallel | `maxSessions` limit reached, sessions rejected, diluted attention | Respect the limit, prioritize, sequence if necessary |\n| Using the agent's `workdir` instead of the project's | Claude Code works in the wrong directory | Always point to the target project directory |\n| Not naming sessions | Hard to identify them in `claude_sessions` | Always use `name` in kebab-case |\n| Auto-responding to critical questions | Decisions made without the user's approval | When in doubt, forward to the user |\n| Ignoring wake events | Sessions stuck waiting indefinitely | Handle each wake event promptly |\n| Not adjusting the budget | Waste (too high) or interruption (too low) | Calibrate `max_budget_usd` based on complexity |\n\n---\n\n## 8. Quick tool reference\n\n| Tool | Usage | Key parameters |\n|---|---|---|\n| `claude_launch` | Launch a session | `prompt`, `name`, `workdir`, `multi_turn`, `max_budget_usd` |\n| `claude_sessions` | List sessions | `status` (all\u002Frunning\u002Fcompleted\u002Ffailed) |\n| `claude_output` | Read the output | `session`, `full`, `lines` |\n| `claude_fg` | Foreground + live stream | `session` |\n| `claude_bg` | Switch back to background | `session` |\n| `claude_kill` | Kill a session | `session` |\n| `claude_respond` | Send a follow-up | `session`, `message`, `interrupt` |\n| `claude_stats` | Usage metrics | none |\n",{"app":50,"name":5,"agents":51,"inputs":53,"attempts":54,"successes":54,"confidence":14,"description":7,"last_adjusted":55,"expected_state":56,"requires_approval":59},"chat-plugin",[52],"*",[],0,"2026-05-11T21:59:41+00:00",{"description":57,"verification_tool":58},"Imported skill — behavior matches source repo","none",true,"master","5f13b5d2b532c0ab6f847e3aa24d4bfe66f7d461","skills\u002Fchat-plugin\u002FClaude Code Orchestration","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os","https:\u002F\u002Fgithub.com\u002FFlissel","https:\u002F\u002Fvibemind.space","Co-founder & CTO, VibeMind - a voice-first agentic OS for non-technical founders. Agents that do the work, not describe it.","2026-11-12 01:01:00",{"source":8,"skillName":5,"commands":69,"prompt":79,"zipAvailable":59},[70,73,76],{"tool":71,"command":72},"npx","npx skills add Flissel\u002Fvibemind-os --skill \"chat-plugin-Claude Code Orchestration\"",{"tool":74,"command":75},"bunx","bunx skills add Flissel\u002Fvibemind-os --skill \"chat-plugin-Claude Code Orchestration\"",{"tool":77,"command":78},"pnpm","pnpm dlx skills add Flissel\u002Fvibemind-os --skill \"chat-plugin-Claude Code Orchestration\"","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fskills\u002Fchat-plugin\u002FClaude Code Orchestration\u002FSKILL.md",[],[82,90,97,125,133,147],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":45},"01M6JAA67M92TRE7GWDJDBTA9T","review","Review a change.","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fspaces\u002Frowboat\u002Frowboat\u002Fpackages\u002Fopenai-plugin-runtime\u002Ftest\u002Ffixtures\u002Fcomplete-plugin\u002Fskills\u002Freview\u002FSKILL.md",[],[84,89],"change",{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":45},"01M6JAA3DZR55G8ZY493BE9NBB","triage","Triage a change.","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fspaces\u002Frowboat\u002Frowboat\u002Fpackages\u002Fopenai-plugin-runtime\u002Ftest\u002Ffixtures\u002Fcomplete-plugin\u002Fskills\u002Fgroup\u002Ftriage\u002FSKILL.md",[],[92,89],{"skillNo":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":101,"categories":102,"keywords":103,"publishedAt":45},"01M6JA9XPVHENK0FQN7HBK54NV","zentrale-werkzeuge","Werkzeuge der Zentrale – nutzen bei GitHub-Fragen (Projekte updaten, Repos, Issues, PRs, Commits, CI\u002FActions-Status, README), bei Fragen zum eigenen Gitea (Sicherung\u002FSpiegel, Repos, Issues, Commits), wenn neueste Docs\u002FDoku zu einer Bibliothek oder einem Dienst geholt werden sollen, wenn ein Dashboard aus meinen Daten gebaut werden soll und wenn ich sage 'auf meinem PC suchen' oder 'such auf dem Rechner nach …'.","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fspaces\u002Frowboat\u002Fdeploy\u002Fheadless\u002Fskills\u002Fzentrale-werkzeuge\u002FSKILL.md",[],[104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124],"zentrale werkzeuge","grep","file","README","list","fetch","todo","gitea","add","URL","mkdir","prs","commits","ci","actions","issues","status","repos","dashboard","github","docs",{"skillNo":126,"name":127,"nameZh":6,"description":128,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":129,"categories":130,"keywords":131,"publishedAt":45},"01M6JA9SQPX187CVC148SQT1SP","wissen-suchen","Bedeutungssuche im Rowboat-Wissen – IMMER zuerst nutzen, wenn nach Kunden, Kontakten, Themen, Interessen, Notizen oder 'was wissen wir über …' gefragt wird, besonders wenn die genauen Wörter unklar sind.","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fspaces\u002Frowboat\u002Fdeploy\u002Fheadless\u002Fskills\u002Fwissen-suchen\u002FSKILL.md",[],[132],"wissen suchen",{"skillNo":134,"name":135,"nameZh":6,"description":136,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":137,"categories":138,"keywords":139,"publishedAt":45},"01M6JA9P6GNFPE21101SDZS3BT","reel-aus-material","Reel aus vorhandenen Laura-Videos: passende Stellen im Transkript finden; spricht dort eine Person, bleibt ihr Originalton (mit Untertiteln aus dem Transkript), sonst trägt ein eingesprochener Satz den Clip. Bei Aufträgen der Art 'material'.","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fspaces\u002Fmarketing\u002Fvideo_claw\u002Fskills\u002Freel-aus-material\u002FSKILL.md",[],[140,141,142,143,144,145,146],"material","reel","reel aus material","person","videos","Art","clip",{"skillNo":148,"name":149,"nameZh":6,"description":150,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":151,"categories":152,"keywords":153,"publishedAt":45},"01M6JA9KSATRT42ACWT66514ZP","reel-aus-bildern","Reel ohne Filmmaterial: Titelkarte, 3–5 Markenbilder mit wechselnder Bewegung, Schlusskarte mit Handlung; Farben und Logo der Firma. Bei Aufträgen der Art 'bilder'.","https:\u002F\u002Fgithub.com\u002FFlissel\u002Fvibemind-os\u002Fblob\u002Fmaster\u002Fspaces\u002Fmarketing\u002Fvideo_claw\u002Fskills\u002Freel-aus-bildern\u002FSKILL.md",[],[141,154,155,145],"reel aus bildern","logo"]