{ "schemaVersion": "1.0", "item": { "slug": "multi-task", "name": "multi task", "source": "tencent", "type": "skill", "category": "开发工具", "sourceUrl": "https://clawhub.ai/brightWeen/multi-task", "canonicalUrl": "https://clawhub.ai/brightWeen/multi-task", "targetPlatform": "OpenClaw" }, "install": { "downloadMode": "redirect", "downloadUrl": "/downloads/multi-task", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=multi-task", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "installMethod": "Manual import", "extraction": "Extract archive", "prerequisites": [ "OpenClaw" ], "packageFormat": "ZIP package", "includedAssets": [ "SKILL.md", "references/advanced-patterns.md" ], "primaryDoc": "SKILL.md", "quickSetup": [ "Download the package from Yavira.", "Extract the archive and review SKILL.md first.", "Import or place the package into your OpenClaw setup." ], "agentAssist": { "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.", "steps": [ "Download the package from Yavira.", "Extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the extracted folder." ], "prompts": [ { "label": "New install", "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete." }, { "label": "Upgrade existing", "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run." } ] }, "sourceHealth": { "source": "tencent", "slug": "multi-task", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-05-12T08:29:19.661Z", "expiresAt": "2026-05-19T08:29:19.661Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=multi-task", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=multi-task", "contentDisposition": "attachment; filename=\"multi-task-1.0.0.zip\"", "redirectLocation": null, "bodySnippet": null, "slug": "multi-task" }, "scope": "item", "summary": "Item download looks usable.", "detail": "Yavira can redirect you to the upstream package for this item.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/multi-task" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] }, "downloadPageUrl": "https://openagent3.xyz/downloads/multi-task", "agentPageUrl": "https://openagent3.xyz/skills/multi-task/agent", "manifestUrl": "https://openagent3.xyz/skills/multi-task/agent.json", "briefUrl": "https://openagent3.xyz/skills/multi-task/agent.md" }, "agentAssist": { "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.", "steps": [ "Download the package from Yavira.", "Extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the extracted folder." ], "prompts": [ { "label": "New install", "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete." }, { "label": "Upgrade existing", "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run." } ] }, "documentation": { "source": "clawhub", "primaryDoc": "SKILL.md", "sections": [ { "title": "Overview", "body": "When users present tasks that decompose into multiple independent work units, serial execution wastes time. This skill guides you to identify batch opportunities, construct self-contained prompts for each unit, and dispatch them as parallel subagents via the Task tool — completing in minutes what would otherwise take much longer sequentially.\n\nThe core insight: a single message can contain multiple Task tool calls, and they all execute concurrently. Your job is to make each subagent's prompt fully self-contained (they cannot see this conversation) and to coordinate the results." }, { "title": "When to Use This Skill", "body": "Strong signals:\n\nUser says \"process all files in X folder\"\nUser provides a list: \"do A, B, C, D for each of these...\"\nUser mentions \"batch\", \"bulk\", \"every\", \"each\", \"all N files\"\nA folder contains multiple files needing the same operation\nUser wants multiple pages/components/reports generated\n\nAlso consider using when:\n\nUser describes repetitive work that you'd otherwise do in a loop\nThe task involves 3+ independent units of similar type\nProcessing time per unit is non-trivial (reading/transforming documents, generating code, etc.)\n\nDo NOT use when:\n\nTasks have strict sequential dependencies (output of task N feeds into task N+1)\nThere are fewer than 3 work units (overhead isn't worth it)\nThe task is a single complex operation that can't be decomposed\nUser explicitly asks for serial processing" }, { "title": "Step 1: ANALYZE — Understand the Work", "body": "Before dispatching anything, enumerate what needs to be done:\n\nList all work units — files to process, pages to build, items to transform\nIdentify the operation — what happens to each unit (extract, convert, summarize, generate, etc.)\nCheck for shared context — do all units need the same template, config, or reference data? If so, read it once now and include it in every subagent prompt.\nDetect dependencies — are any units dependent on others? If yes, read references/advanced-patterns.md for dependency handling strategies. If all units are independent (the common case), proceed directly.\nCount the units — this determines your dispatch strategy:\n\n3-10 units: single wave, all parallel\n11-50 units: dispatch in waves of 8-10\n50+ units: run 2-3 pilot tasks first to validate your prompt, then dispatch the rest in waves\n\nPresent your analysis to the user:\n\nFound N work units: [brief list]\nOperation: [what will happen to each]\nShared context: [any common dependencies]\nDependencies: [none / description]\nStrategy: [single wave / M waves of ~K / pilot + waves]\n\nWait for user confirmation before dispatching, especially for large batches." }, { "title": "Step 2: PLAN — Decompose into Task Units", "body": "For each work unit, define:\n\ntask-ID: Sequential identifier (task-001, task-002, ...)\ninput: Absolute path(s) to input file(s) or data\noperation: What the subagent should do\noutput: Absolute path for results (ensure no path conflicts between tasks)\nrecommended skill: If the task matches an installed skill (see Skill Matching below)\n\nOutput path strategy: Create a dedicated output directory to keep results organized:\n\n/multi-task-output/\n├── task-001/\n├── task-002/\n└── ...\n\nUse mkdir -p to create the output directory structure before dispatching." }, { "title": "Step 3: PROMPT — Construct Subagent Prompts", "body": "Each subagent starts with a blank context — it cannot see this conversation. Every prompt must be completely self-contained. Use this template:\n\n## Task [task-ID]: [Brief description]\n\n### Skill Recommendation\n[If a matching skill is available]:\nYou have access to the `/[skill-name]` skill which is ideal for this task.\nInvoke it using the Skill tool with skill=\"[skill-name]\" to get specialized\ninstructions before proceeding.\n\n### Context\n[Any shared context the subagent needs — project background, conventions,\ntemplates, reference data. Include the actual content, not references to\n\"the conversation above\".]\n\n### Input\n- File: [absolute path]\n- [Any other inputs, with absolute paths]\n\n### Instructions\n[Clear, step-by-step instructions for what to do]\n1. [Step 1]\n2. [Step 2]\n...\n\n### Output\n- Save results to: [absolute path to task-specific output directory]\n- Expected deliverables: [list of output files]\n- [Any format requirements]\n\n### Important Notes\n- Use absolute paths for all file operations\n- Do not modify the input file(s)\n- If you encounter an error, save error details to [output-dir]/error.log\n\nPrompt quality checklist:\n\nAll paths are absolute\n No references to \"the conversation\" or \"as discussed\"\n Shared context is included verbatim, not by reference\n Output paths are unique per task (no conflicts)\n Instructions are specific enough for an agent with no prior context\n Skill recommendation is included if applicable" }, { "title": "Step 4: DISPATCH — Send Tasks in Parallel", "body": "The critical mechanism: Include multiple Task tool calls in a single message. This is what makes them parallel. If you send them in separate messages, they run serially.\n\nFor 3-10 tasks: Send all in one message:\n\n[Single message containing:]\nTask(subagent_type=\"general-purpose\", prompt=\"## Task task-001: ...\", description=\"Process file-001\")\nTask(subagent_type=\"general-purpose\", prompt=\"## Task task-002: ...\", description=\"Process file-002\")\nTask(subagent_type=\"general-purpose\", prompt=\"## Task task-003: ...\", description=\"Process file-003\")\n...\n\nFor 11-50 tasks: Dispatch in waves of 8-10. Wait for each wave to complete before starting the next:\n\nWave 1: task-001 through task-010 (single message, all parallel)\n[Wait for completion, report progress]\nWave 2: task-011 through task-020 (single message, all parallel)\n[Wait for completion, report progress]\n...\n\nFor 50+ tasks: Run a pilot first:\n\nPick 2-3 representative tasks (include edge cases if possible)\nDispatch them as a pilot wave\nVerify results are correct\nIf issues found, fix the prompt template and re-pilot\nOnce validated, dispatch the remaining tasks in waves of 8-10\n\nSubagent type selection:\n\nDefault: general-purpose (has access to all tools including Skill)\nFor pure shell/git tasks: Bash\nFor code exploration only: Explore\n\nRun tasks in background when appropriate: For large batches, use run_in_background: true so you can monitor progress and report to the user incrementally." }, { "title": "Step 5: MONITOR — Track Progress", "body": "As results come back:\n\nTrack completion — maintain a mental tally: \"Wave 1: 8/10 complete\"\nCheck for failures — if a task fails:\n\nAnalyze the error\nFix the prompt if needed\nRetry up to 2 times automatically\nIf still failing after 2 retries, mark as failed and continue with others\n\n\nReport progress to user after each wave:\nWave 1 complete: 9/10 succeeded, 1 failed (task-007: [reason])\nStarting wave 2...\n\n\nFailure isolation — one task's failure must never block or affect other tasks" }, { "title": "Step 6: MERGE — Collect and Present Results", "body": "After all waves complete:\n\nSort results by task-ID — present in order regardless of completion time\nSummarize outcomes:\nBatch complete: N/M tasks succeeded\n\nSuccessful:\n- task-001: [output path] — [brief description]\n- task-002: [output path] — [brief description]\n...\n\nFailed (if any):\n- task-007: [error reason] — [suggested fix]\n\n\nHandle failed tasks — offer to retry failed tasks or let user fix input and rerun\nMerge outputs if requested — some batch operations need a final merge step (e.g., combining extracted text into one document). Do this after all tasks complete." }, { "title": "Skill Matching", "body": "When planning tasks, match each work unit against installed skills. This dramatically improves subagent performance because skills provide specialized, tested instructions.\n\nMatching rules:\n\nTask involves...Recommend skillPDF files (read, create, merge, extract)/pdfWord documents (.docx read, create, edit)/docxPowerPoint files (.pptx)/pptxSpreadsheets (.xlsx, .csv, .tsv)/xlsxWeb pages, components, HTML/CSS/frontend-designVisual design, posters, art/canvas-designStyling artifacts with themes/theme-factory\n\nHow to include skill recommendations in prompts:\n\nIn each subagent's prompt, add the skill invocation instruction:\n\n### Skill Recommendation\nYou have access to the `/pdf` skill. Before starting work, invoke it using\nthe Skill tool: Skill(skill=\"pdf\"). This will load specialized instructions\nfor PDF processing that will help you complete this task more effectively.\n\nIf no installed skill matches, omit the skill recommendation section — the subagent will use its general capabilities." }, { "title": "Example 1: Batch PDF Text Extraction", "body": "User: \"Extract text from all PDFs in /Users/me/reports/ and save as markdown files\"\n\nAnalysis:\n\nFound 12 PDF files in /Users/me/reports/\nOperation: Extract text from each PDF, save as .md\nShared context: None\nDependencies: None\nStrategy: 2 waves of 6\n\nSubagent prompt (each task):\n\n## Task task-001: Extract text from Q1-report.pdf\n\n### Skill Recommendation\nYou have access to the `/pdf` skill. Invoke it using the Skill tool with\nskill=\"pdf\" to get specialized PDF processing instructions.\n\n### Input\n- File: /Users/me/reports/Q1-report.pdf\n\n### Instructions\n1. Read the PDF file and extract all text content\n2. Preserve heading structure where possible\n3. Format the output as clean Markdown\n4. Include page breaks as horizontal rules (---)\n\n### Output\n- Save to: /Users/me/reports/multi-task-output/task-001/Q1-report.md\n- Create the output directory if it doesn't exist" }, { "title": "Example 2: Multi-Page Frontend Development", "body": "User: \"Build 5 pages for our marketing site: Home, About, Pricing, Blog, Contact\"\n\nAnalysis:\n\nFound 5 work units: Home, About, Pricing, Blog, Contact pages\nOperation: Generate frontend code for each page\nShared context: Brand guidelines, shared layout components, color scheme\nDependencies: None (each page is independent)\nStrategy: Single wave, all 5 parallel\n\nKey considerations:\n\nRead any existing shared components/styles first\nInclude the full shared context (brand colors, fonts, layout patterns) in each prompt\nEach subagent gets /frontend-design skill recommendation\nOutput to separate directories: pages/home/, pages/about/, etc." }, { "title": "Example 3: Batch Data Conversion", "body": "User: \"Convert all CSV files in /data/raw/ to JSON format with proper types\"\n\nAnalysis:\n\nFound 25 CSV files in /data/raw/\nOperation: Parse CSV, infer types, convert to JSON\nShared context: Type inference rules (dates, numbers, booleans)\nDependencies: None\nStrategy: 3 waves of ~8-9\n\nKey considerations:\n\nEach subagent gets /xlsx skill recommendation (handles CSV)\nInclude type inference rules in every prompt\nStandardize output format in the prompt template" }, { "title": "Troubleshooting", "body": "ProblemCauseFixTasks run serially, not parallelTask calls sent in separate messagesPut ALL Task calls in a single messageSubagent says \"I don't have context\"Prompt references conversation historyMake prompt fully self-containedFile not found errorsRelative paths usedUse absolute paths everywhereOutput files overwrite each otherSame output path for multiple tasksUse task-ID in output directory pathSubagent doesn't use recommended skillSkill instruction unclearAdd explicit Skill(skill=\"name\") call instructionToo many tasks overwhelm the systemDispatching 50+ at onceUse waves of 8-10Results arrive in wrong orderRelying on completion orderSort by task-ID, not completion timeOne failure cascades to othersShared state between tasksEnsure full isolation — separate dirs, no shared files" }, { "title": "Advanced Patterns", "body": "For handling tasks with dependencies (linear chains, fan-in/fan-out, partial dependencies), dynamic batch sizing, and conditional dispatch, read references/advanced-patterns.md." } ], "body": "Multi-Task: Parallel Batch Orchestration\nOverview\n\nWhen users present tasks that decompose into multiple independent work units, serial execution wastes time. This skill guides you to identify batch opportunities, construct self-contained prompts for each unit, and dispatch them as parallel subagents via the Task tool — completing in minutes what would otherwise take much longer sequentially.\n\nThe core insight: a single message can contain multiple Task tool calls, and they all execute concurrently. Your job is to make each subagent's prompt fully self-contained (they cannot see this conversation) and to coordinate the results.\n\nWhen to Use This Skill\n\nStrong signals:\n\nUser says \"process all files in X folder\"\nUser provides a list: \"do A, B, C, D for each of these...\"\nUser mentions \"batch\", \"bulk\", \"every\", \"each\", \"all N files\"\nA folder contains multiple files needing the same operation\nUser wants multiple pages/components/reports generated\n\nAlso consider using when:\n\nUser describes repetitive work that you'd otherwise do in a loop\nThe task involves 3+ independent units of similar type\nProcessing time per unit is non-trivial (reading/transforming documents, generating code, etc.)\n\nDo NOT use when:\n\nTasks have strict sequential dependencies (output of task N feeds into task N+1)\nThere are fewer than 3 work units (overhead isn't worth it)\nThe task is a single complex operation that can't be decomposed\nUser explicitly asks for serial processing\nThe 6-Step Workflow\nStep 1: ANALYZE — Understand the Work\n\nBefore dispatching anything, enumerate what needs to be done:\n\nList all work units — files to process, pages to build, items to transform\nIdentify the operation — what happens to each unit (extract, convert, summarize, generate, etc.)\nCheck for shared context — do all units need the same template, config, or reference data? If so, read it once now and include it in every subagent prompt.\nDetect dependencies — are any units dependent on others? If yes, read references/advanced-patterns.md for dependency handling strategies. If all units are independent (the common case), proceed directly.\nCount the units — this determines your dispatch strategy:\n3-10 units: single wave, all parallel\n11-50 units: dispatch in waves of 8-10\n50+ units: run 2-3 pilot tasks first to validate your prompt, then dispatch the rest in waves\n\nPresent your analysis to the user:\n\nFound N work units: [brief list]\nOperation: [what will happen to each]\nShared context: [any common dependencies]\nDependencies: [none / description]\nStrategy: [single wave / M waves of ~K / pilot + waves]\n\n\nWait for user confirmation before dispatching, especially for large batches.\n\nStep 2: PLAN — Decompose into Task Units\n\nFor each work unit, define:\n\ntask-ID: Sequential identifier (task-001, task-002, ...)\ninput: Absolute path(s) to input file(s) or data\noperation: What the subagent should do\noutput: Absolute path for results (ensure no path conflicts between tasks)\nrecommended skill: If the task matches an installed skill (see Skill Matching below)\n\nOutput path strategy: Create a dedicated output directory to keep results organized:\n\n/multi-task-output/\n├── task-001/\n├── task-002/\n└── ...\n\n\nUse mkdir -p to create the output directory structure before dispatching.\n\nStep 3: PROMPT — Construct Subagent Prompts\n\nEach subagent starts with a blank context — it cannot see this conversation. Every prompt must be completely self-contained. Use this template:\n\n## Task [task-ID]: [Brief description]\n\n### Skill Recommendation\n[If a matching skill is available]:\nYou have access to the `/[skill-name]` skill which is ideal for this task.\nInvoke it using the Skill tool with skill=\"[skill-name]\" to get specialized\ninstructions before proceeding.\n\n### Context\n[Any shared context the subagent needs — project background, conventions,\ntemplates, reference data. Include the actual content, not references to\n\"the conversation above\".]\n\n### Input\n- File: [absolute path]\n- [Any other inputs, with absolute paths]\n\n### Instructions\n[Clear, step-by-step instructions for what to do]\n1. [Step 1]\n2. [Step 2]\n...\n\n### Output\n- Save results to: [absolute path to task-specific output directory]\n- Expected deliverables: [list of output files]\n- [Any format requirements]\n\n### Important Notes\n- Use absolute paths for all file operations\n- Do not modify the input file(s)\n- If you encounter an error, save error details to [output-dir]/error.log\n\n\nPrompt quality checklist:\n\n All paths are absolute\n No references to \"the conversation\" or \"as discussed\"\n Shared context is included verbatim, not by reference\n Output paths are unique per task (no conflicts)\n Instructions are specific enough for an agent with no prior context\n Skill recommendation is included if applicable\nStep 4: DISPATCH — Send Tasks in Parallel\n\nThe critical mechanism: Include multiple Task tool calls in a single message. This is what makes them parallel. If you send them in separate messages, they run serially.\n\nFor 3-10 tasks: Send all in one message:\n\n[Single message containing:]\nTask(subagent_type=\"general-purpose\", prompt=\"## Task task-001: ...\", description=\"Process file-001\")\nTask(subagent_type=\"general-purpose\", prompt=\"## Task task-002: ...\", description=\"Process file-002\")\nTask(subagent_type=\"general-purpose\", prompt=\"## Task task-003: ...\", description=\"Process file-003\")\n...\n\n\nFor 11-50 tasks: Dispatch in waves of 8-10. Wait for each wave to complete before starting the next:\n\nWave 1: task-001 through task-010 (single message, all parallel)\n[Wait for completion, report progress]\nWave 2: task-011 through task-020 (single message, all parallel)\n[Wait for completion, report progress]\n...\n\n\nFor 50+ tasks: Run a pilot first:\n\nPick 2-3 representative tasks (include edge cases if possible)\nDispatch them as a pilot wave\nVerify results are correct\nIf issues found, fix the prompt template and re-pilot\nOnce validated, dispatch the remaining tasks in waves of 8-10\n\nSubagent type selection:\n\nDefault: general-purpose (has access to all tools including Skill)\nFor pure shell/git tasks: Bash\nFor code exploration only: Explore\n\nRun tasks in background when appropriate: For large batches, use run_in_background: true so you can monitor progress and report to the user incrementally.\n\nStep 5: MONITOR — Track Progress\n\nAs results come back:\n\nTrack completion — maintain a mental tally: \"Wave 1: 8/10 complete\"\nCheck for failures — if a task fails:\nAnalyze the error\nFix the prompt if needed\nRetry up to 2 times automatically\nIf still failing after 2 retries, mark as failed and continue with others\nReport progress to user after each wave:\nWave 1 complete: 9/10 succeeded, 1 failed (task-007: [reason])\nStarting wave 2...\n\nFailure isolation — one task's failure must never block or affect other tasks\nStep 6: MERGE — Collect and Present Results\n\nAfter all waves complete:\n\nSort results by task-ID — present in order regardless of completion time\nSummarize outcomes:\nBatch complete: N/M tasks succeeded\n\nSuccessful:\n- task-001: [output path] — [brief description]\n- task-002: [output path] — [brief description]\n...\n\nFailed (if any):\n- task-007: [error reason] — [suggested fix]\n\nHandle failed tasks — offer to retry failed tasks or let user fix input and rerun\nMerge outputs if requested — some batch operations need a final merge step (e.g., combining extracted text into one document). Do this after all tasks complete.\nSkill Matching\n\nWhen planning tasks, match each work unit against installed skills. This dramatically improves subagent performance because skills provide specialized, tested instructions.\n\nMatching rules:\n\nTask involves...\tRecommend skill\nPDF files (read, create, merge, extract)\t/pdf\nWord documents (.docx read, create, edit)\t/docx\nPowerPoint files (.pptx)\t/pptx\nSpreadsheets (.xlsx, .csv, .tsv)\t/xlsx\nWeb pages, components, HTML/CSS\t/frontend-design\nVisual design, posters, art\t/canvas-design\nStyling artifacts with themes\t/theme-factory\n\nHow to include skill recommendations in prompts:\n\nIn each subagent's prompt, add the skill invocation instruction:\n\n### Skill Recommendation\nYou have access to the `/pdf` skill. Before starting work, invoke it using\nthe Skill tool: Skill(skill=\"pdf\"). This will load specialized instructions\nfor PDF processing that will help you complete this task more effectively.\n\n\nIf no installed skill matches, omit the skill recommendation section — the subagent will use its general capabilities.\n\nExamples\nExample 1: Batch PDF Text Extraction\n\nUser: \"Extract text from all PDFs in /Users/me/reports/ and save as markdown files\"\n\nAnalysis:\n\nFound 12 PDF files in /Users/me/reports/\nOperation: Extract text from each PDF, save as .md\nShared context: None\nDependencies: None\nStrategy: 2 waves of 6\n\n\nSubagent prompt (each task):\n\n## Task task-001: Extract text from Q1-report.pdf\n\n### Skill Recommendation\nYou have access to the `/pdf` skill. Invoke it using the Skill tool with\nskill=\"pdf\" to get specialized PDF processing instructions.\n\n### Input\n- File: /Users/me/reports/Q1-report.pdf\n\n### Instructions\n1. Read the PDF file and extract all text content\n2. Preserve heading structure where possible\n3. Format the output as clean Markdown\n4. Include page breaks as horizontal rules (---)\n\n### Output\n- Save to: /Users/me/reports/multi-task-output/task-001/Q1-report.md\n- Create the output directory if it doesn't exist\n\nExample 2: Multi-Page Frontend Development\n\nUser: \"Build 5 pages for our marketing site: Home, About, Pricing, Blog, Contact\"\n\nAnalysis:\n\nFound 5 work units: Home, About, Pricing, Blog, Contact pages\nOperation: Generate frontend code for each page\nShared context: Brand guidelines, shared layout components, color scheme\nDependencies: None (each page is independent)\nStrategy: Single wave, all 5 parallel\n\n\nKey considerations:\n\nRead any existing shared components/styles first\nInclude the full shared context (brand colors, fonts, layout patterns) in each prompt\nEach subagent gets /frontend-design skill recommendation\nOutput to separate directories: pages/home/, pages/about/, etc.\nExample 3: Batch Data Conversion\n\nUser: \"Convert all CSV files in /data/raw/ to JSON format with proper types\"\n\nAnalysis:\n\nFound 25 CSV files in /data/raw/\nOperation: Parse CSV, infer types, convert to JSON\nShared context: Type inference rules (dates, numbers, booleans)\nDependencies: None\nStrategy: 3 waves of ~8-9\n\n\nKey considerations:\n\nEach subagent gets /xlsx skill recommendation (handles CSV)\nInclude type inference rules in every prompt\nStandardize output format in the prompt template\nTroubleshooting\nProblem\tCause\tFix\nTasks run serially, not parallel\tTask calls sent in separate messages\tPut ALL Task calls in a single message\nSubagent says \"I don't have context\"\tPrompt references conversation history\tMake prompt fully self-contained\nFile not found errors\tRelative paths used\tUse absolute paths everywhere\nOutput files overwrite each other\tSame output path for multiple tasks\tUse task-ID in output directory path\nSubagent doesn't use recommended skill\tSkill instruction unclear\tAdd explicit Skill(skill=\"name\") call instruction\nToo many tasks overwhelm the system\tDispatching 50+ at once\tUse waves of 8-10\nResults arrive in wrong order\tRelying on completion order\tSort by task-ID, not completion time\nOne failure cascades to others\tShared state between tasks\tEnsure full isolation — separate dirs, no shared files\nAdvanced Patterns\n\nFor handling tasks with dependencies (linear chains, fan-in/fan-out, partial dependencies), dynamic batch sizing, and conditional dispatch, read references/advanced-patterns.md." }, "trust": { "sourceLabel": "tencent", "provenanceUrl": "https://clawhub.ai/brightWeen/multi-task", "publisherUrl": "https://clawhub.ai/brightWeen/multi-task", "owner": "brightWeen", "version": "1.0.0", "license": null, "verificationStatus": "Indexed source record" }, "links": { "detailUrl": "https://openagent3.xyz/skills/multi-task", "downloadUrl": "https://openagent3.xyz/downloads/multi-task", "agentUrl": "https://openagent3.xyz/skills/multi-task/agent", "manifestUrl": "https://openagent3.xyz/skills/multi-task/agent.json", "briefUrl": "https://openagent3.xyz/skills/multi-task/agent.md" } }