{ "schemaVersion": "1.0", "item": { "slug": "subagent-driven-development-2", "name": "Subagent Driven Development", "source": "tencent", "type": "skill", "category": "效率提升", "sourceUrl": "https://clawhub.ai/wpank/subagent-driven-development-2", "canonicalUrl": "https://clawhub.ai/wpank/subagent-driven-development-2", "targetPlatform": "OpenClaw" }, "install": { "downloadMode": "redirect", "downloadUrl": "/downloads/subagent-driven-development-2", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=subagent-driven-development-2", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "installMethod": "Manual import", "extraction": "Extract archive", "prerequisites": [ "OpenClaw" ], "packageFormat": "ZIP package", "includedAssets": [ "SKILL.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", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-04-30T16:55:25.780Z", "expiresAt": "2026-05-07T16:55:25.780Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentDisposition": "attachment; filename=\"network-1.0.0.zip\"", "redirectLocation": null, "bodySnippet": null }, "scope": "source", "summary": "Source download looks usable.", "detail": "Yavira can redirect you to the upstream package for this source.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/subagent-driven-development-2" }, "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/subagent-driven-development-2", "agentPageUrl": "https://openagent3.xyz/skills/subagent-driven-development-2/agent", "manifestUrl": "https://openagent3.xyz/skills/subagent-driven-development-2/agent.json", "briefUrl": "https://openagent3.xyz/skills/subagent-driven-development-2/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": "Subagent-Driven Development", "body": "Execute a plan by dispatching a fresh subagent per task, with two-stage review after each — spec compliance first, then code quality.\n\nCore principle: Fresh subagent per task + two-stage review = no context pollution, high quality, fast iteration." }, { "title": "OpenClaw / Moltbot / Clawbot", "body": "npx clawhub@latest install subagent-driven-development" }, { "title": "When to Use This Skill", "body": "You have an implementation plan with discrete, mostly independent tasks\nYou want to execute the plan within a single session (no human-in-the-loop between tasks)\nTasks can be implemented and reviewed sequentially without tight coupling\n\nConditionThis SkillAlternativeHave a plan, same sessionYes—Have a plan, parallel sessionNoexecuting-plansNo plan yetNoWrite a plan firstTasks tightly coupledNoManual execution or decompose more" }, { "title": "The Process", "body": "┌─────────────────────────────────────────────────────┐\n│ 1. Read plan, extract ALL tasks with full text │\n│ 2. Note shared context (arch, deps, conventions) │\n│ 3. Create TodoWrite with all tasks │\n└──────────────────────┬──────────────────────────────┘\n ▼\n ┌────── Per Task ──────┐\n │ │\n │ Dispatch Implementer (references/implementer-prompt.md)\n │ │ │\n │ ▼ │\n │ Questions?──yes──► Answer, re-dispatch│\n │ │no │\n │ ▼ │\n │ Implement + test + commit + self-review│\n │ │ │\n │ ▼ │\n │ Dispatch Spec Reviewer (references/spec-reviewer-prompt.md)\n │ │ │\n │ ▼ │\n │ Compliant?──no──► Implementer fixes──┘\n │ │yes then re-review\n │ ▼\n │ Dispatch Code Reviewer (references/code-quality-reviewer-prompt.md)\n │ │\n │ ▼\n │ Approved?──no──► Implementer fixes, re-review\n │ │yes\n │ ▼\n │ Mark task complete\n └───────┬──────────────┘\n ▼\n More tasks? ──yes──► next task\n │no\n ▼\n Final cross-task code review\n ▼\n Finish development branch" }, { "title": "Prompt Templates", "body": "Three reference prompts are provided for the subagent roles:\n\nRoleFilePurposeImplementerreferences/implementer-prompt.mdImplement, test, commit, self-reviewSpec reviewerreferences/spec-reviewer-prompt.mdVerify code matches spec exactlyCode reviewerreferences/code-quality-reviewer-prompt.mdVerify code is clean and maintainable" }, { "title": "Example Workflow", "body": "Controller: Reading plan — 5 tasks extracted, TodoWrite created.\n\n─── Task 1: Hook installation script ───\n\n[Dispatch implementer with full task text + context]\n\nImplementer: \"Should the hook be installed at user or system level?\"\nController: \"User level (~/.config/hooks/)\"\n\nImplementer: ✅ Implemented install-hook command\n - Added tests (5/5 passing)\n - Self-review: missed --force flag, added it\n - Committed\n\n[Dispatch spec reviewer]\nSpec reviewer: ✅ Spec compliant — all requirements met\n\n[Dispatch code reviewer with git SHAs]\nCode reviewer: ✅ Approved — clean, good coverage\n\n[Mark Task 1 complete]\n\n─── Task 2: Recovery modes ───\n\n[Dispatch implementer]\n\nImplementer: ✅ Added verify/repair modes (8/8 tests passing)\n\n[Dispatch spec reviewer]\nSpec reviewer: ❌ Issues:\n - Missing: progress reporting (\"report every 100 items\")\n - Extra: added --json flag (not in spec)\n\n[Implementer fixes: remove --json, add progress reporting]\nSpec reviewer: ✅ Spec compliant\n\n[Dispatch code reviewer]\nCode reviewer: Important — magic number 100, extract constant\n\n[Implementer fixes: extract PROGRESS_INTERVAL]\nCode reviewer: ✅ Approved\n\n[Mark Task 2 complete]\n\n... (tasks 3-5) ...\n\n[Final cross-task code review]\nFinal reviewer: ✅ All requirements met, ready to merge" }, { "title": "Controller Responsibilities", "body": "The controller (you) orchestrates the flow. Key duties:\n\nResponsibilityDetailExtract tasks upfrontRead plan once, extract all task text — subagents never read the plan fileProvide full contextGive each subagent the complete task text + architectural contextAnswer questionsRespond clearly and completely before letting subagent proceedEnforce review orderSpec compliance first, code quality second — never reversedTrack progressUpdate TodoWrite after each task completesDispatch sequentiallyOne implementation subagent at a time to avoid conflicts" }, { "title": "Quality Gates", "body": "Each task passes through three quality checks:\n\nGateWhoWhatSelf-reviewImplementerCompleteness, naming, YAGNI, test qualitySpec reviewSpec reviewerMatches requirements exactly — nothing more, lessCode reviewCode reviewerClean code, maintainability, test coverage" }, { "title": "Advantages", "body": "vs. manual execution:\n\nFresh context per task (no confusion from accumulated state)\nSubagents follow TDD naturally\nQuestions surfaced before work begins, not after\n\nvs. parallel-session plans:\n\nNo handoff overhead\nContinuous progress\nReview checkpoints are automatic\n\nCost trade-off: More subagent invocations (implementer + 2 reviewers per task), but catches issues early — cheaper than debugging later." }, { "title": "NEVER Do", "body": "Skip either review — both spec compliance and code quality are mandatory\nStart code review before spec review passes — wrong order wastes effort on code that doesn't meet spec\nDispatch multiple implementers in parallel — causes merge conflicts and context corruption\nMake subagents read the plan file — provide full task text directly\nSkip scene-setting context — subagent needs to understand where the task fits architecturally\nIgnore subagent questions — answer fully before they proceed\nAccept \"close enough\" on spec compliance — if the reviewer found issues, it's not done\nSkip re-review after fixes — reviewer found issues → implementer fixes → reviewer reviews again\nLet self-review replace actual review — self-review is a first pass, not a substitute\nMove to next task with open review issues — current task must be fully approved first\nFix issues manually instead of through a subagent — manual fixes pollute controller context" }, { "title": "Handling Failures", "body": "SituationResponseSubagent asks questionsAnswer clearly, provide additional context if neededReviewer finds issuesSame implementer subagent fixes, reviewer re-reviewsSubagent fails task entirelyDispatch a new fix subagent with specific instructionsTask blocked by dependencyReorder remaining tasks or resolve dependency first" } ], "body": "Subagent-Driven Development\n\nExecute a plan by dispatching a fresh subagent per task, with two-stage review after each — spec compliance first, then code quality.\n\nCore principle: Fresh subagent per task + two-stage review = no context pollution, high quality, fast iteration.\n\nInstallation\nOpenClaw / Moltbot / Clawbot\nnpx clawhub@latest install subagent-driven-development\n\nWhen to Use This Skill\nYou have an implementation plan with discrete, mostly independent tasks\nYou want to execute the plan within a single session (no human-in-the-loop between tasks)\nTasks can be implemented and reviewed sequentially without tight coupling\nCondition\tThis Skill\tAlternative\nHave a plan, same session\tYes\t—\nHave a plan, parallel session\tNo\texecuting-plans\nNo plan yet\tNo\tWrite a plan first\nTasks tightly coupled\tNo\tManual execution or decompose more\nThe Process\n┌─────────────────────────────────────────────────────┐\n│ 1. Read plan, extract ALL tasks with full text │\n│ 2. Note shared context (arch, deps, conventions) │\n│ 3. Create TodoWrite with all tasks │\n└──────────────────────┬──────────────────────────────┘\n ▼\n ┌────── Per Task ──────┐\n │ │\n │ Dispatch Implementer (references/implementer-prompt.md)\n │ │ │\n │ ▼ │\n │ Questions?──yes──► Answer, re-dispatch│\n │ │no │\n │ ▼ │\n │ Implement + test + commit + self-review│\n │ │ │\n │ ▼ │\n │ Dispatch Spec Reviewer (references/spec-reviewer-prompt.md)\n │ │ │\n │ ▼ │\n │ Compliant?──no──► Implementer fixes──┘\n │ │yes then re-review\n │ ▼\n │ Dispatch Code Reviewer (references/code-quality-reviewer-prompt.md)\n │ │\n │ ▼\n │ Approved?──no──► Implementer fixes, re-review\n │ │yes\n │ ▼\n │ Mark task complete\n └───────┬──────────────┘\n ▼\n More tasks? ──yes──► next task\n │no\n ▼\n Final cross-task code review\n ▼\n Finish development branch\n\nPrompt Templates\n\nThree reference prompts are provided for the subagent roles:\n\nRole\tFile\tPurpose\nImplementer\treferences/implementer-prompt.md\tImplement, test, commit, self-review\nSpec reviewer\treferences/spec-reviewer-prompt.md\tVerify code matches spec exactly\nCode reviewer\treferences/code-quality-reviewer-prompt.md\tVerify code is clean and maintainable\nExample Workflow\nController: Reading plan — 5 tasks extracted, TodoWrite created.\n\n─── Task 1: Hook installation script ───\n\n[Dispatch implementer with full task text + context]\n\nImplementer: \"Should the hook be installed at user or system level?\"\nController: \"User level (~/.config/hooks/)\"\n\nImplementer: ✅ Implemented install-hook command\n - Added tests (5/5 passing)\n - Self-review: missed --force flag, added it\n - Committed\n\n[Dispatch spec reviewer]\nSpec reviewer: ✅ Spec compliant — all requirements met\n\n[Dispatch code reviewer with git SHAs]\nCode reviewer: ✅ Approved — clean, good coverage\n\n[Mark Task 1 complete]\n\n─── Task 2: Recovery modes ───\n\n[Dispatch implementer]\n\nImplementer: ✅ Added verify/repair modes (8/8 tests passing)\n\n[Dispatch spec reviewer]\nSpec reviewer: ❌ Issues:\n - Missing: progress reporting (\"report every 100 items\")\n - Extra: added --json flag (not in spec)\n\n[Implementer fixes: remove --json, add progress reporting]\nSpec reviewer: ✅ Spec compliant\n\n[Dispatch code reviewer]\nCode reviewer: Important — magic number 100, extract constant\n\n[Implementer fixes: extract PROGRESS_INTERVAL]\nCode reviewer: ✅ Approved\n\n[Mark Task 2 complete]\n\n... (tasks 3-5) ...\n\n[Final cross-task code review]\nFinal reviewer: ✅ All requirements met, ready to merge\n\nController Responsibilities\n\nThe controller (you) orchestrates the flow. Key duties:\n\nResponsibility\tDetail\nExtract tasks upfront\tRead plan once, extract all task text — subagents never read the plan file\nProvide full context\tGive each subagent the complete task text + architectural context\nAnswer questions\tRespond clearly and completely before letting subagent proceed\nEnforce review order\tSpec compliance first, code quality second — never reversed\nTrack progress\tUpdate TodoWrite after each task completes\nDispatch sequentially\tOne implementation subagent at a time to avoid conflicts\nQuality Gates\n\nEach task passes through three quality checks:\n\nGate\tWho\tWhat\nSelf-review\tImplementer\tCompleteness, naming, YAGNI, test quality\nSpec review\tSpec reviewer\tMatches requirements exactly — nothing more, less\nCode review\tCode reviewer\tClean code, maintainability, test coverage\nAdvantages\n\nvs. manual execution:\n\nFresh context per task (no confusion from accumulated state)\nSubagents follow TDD naturally\nQuestions surfaced before work begins, not after\n\nvs. parallel-session plans:\n\nNo handoff overhead\nContinuous progress\nReview checkpoints are automatic\n\nCost trade-off: More subagent invocations (implementer + 2 reviewers per task), but catches issues early — cheaper than debugging later.\n\nNEVER Do\nSkip either review — both spec compliance and code quality are mandatory\nStart code review before spec review passes — wrong order wastes effort on code that doesn't meet spec\nDispatch multiple implementers in parallel — causes merge conflicts and context corruption\nMake subagents read the plan file — provide full task text directly\nSkip scene-setting context — subagent needs to understand where the task fits architecturally\nIgnore subagent questions — answer fully before they proceed\nAccept \"close enough\" on spec compliance — if the reviewer found issues, it's not done\nSkip re-review after fixes — reviewer found issues → implementer fixes → reviewer reviews again\nLet self-review replace actual review — self-review is a first pass, not a substitute\nMove to next task with open review issues — current task must be fully approved first\nFix issues manually instead of through a subagent — manual fixes pollute controller context\nHandling Failures\nSituation\tResponse\nSubagent asks questions\tAnswer clearly, provide additional context if needed\nReviewer finds issues\tSame implementer subagent fixes, reviewer re-reviews\nSubagent fails task entirely\tDispatch a new fix subagent with specific instructions\nTask blocked by dependency\tReorder remaining tasks or resolve dependency first" }, "trust": { "sourceLabel": "tencent", "provenanceUrl": "https://clawhub.ai/wpank/subagent-driven-development-2", "publisherUrl": "https://clawhub.ai/wpank/subagent-driven-development-2", "owner": "wpank", "version": "0.1.0", "license": null, "verificationStatus": "Indexed source record" }, "links": { "detailUrl": "https://openagent3.xyz/skills/subagent-driven-development-2", "downloadUrl": "https://openagent3.xyz/downloads/subagent-driven-development-2", "agentUrl": "https://openagent3.xyz/skills/subagent-driven-development-2/agent", "manifestUrl": "https://openagent3.xyz/skills/subagent-driven-development-2/agent.json", "briefUrl": "https://openagent3.xyz/skills/subagent-driven-development-2/agent.md" } }