[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6CECFX9GPFXNGPV31EGRG3V":3,"skill-same:01M6CECFX9GPFXNGPV31EGRG3V":107},{"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":28,"publishedAt":49,"linkOnly":50,"markdownContent":51,"markdownBody":52,"frontMatter":53,"refName":55,"commitSha":56,"skillDirectory":57,"forksCount":58,"repositoryHtmlUrl":59,"repositoryHomepageUrl":6,"ownerHtmlUrl":60,"ownerBlogUrl":61,"ownerDescription":6,"updatedAt":62,"install":63,"resources":76},"01M6CECFX9GPFXNGPV31EGRG3V","agent-development",null,"Design and build custom Claude Code agents with effective descriptions, tool access patterns,\nand self-documenting prompts. Covers Task tool delegation, model selection, memory limits,\nand declarative instruction design.\n\nUse when: creating custom agents, designing agent descriptions for auto-delegation,\ntroubleshooting agent memory issues, or building agent pipelines.","sundial-org\u002Fawesome-openclaw-skills","sundial-org","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F225963689?v=4","01M6CEAPGV2C2CVE8A0VCS0XZH","Sundial","01M6CEAPN7KM9TGZMBC07NDY3H",674,"MIT","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48],"development","agent development","build","design","pipelines","prompts","auto","use","model","troubleshooting","tool","access","custom agents","delegation","issues","agent memory","code agents","code","model selection","memory","2026-11-10 10:12:30",false,"---\nname: agent-development\ndescription: |\n  Design and build custom Claude Code agents with effective descriptions, tool access patterns,\n  and self-documenting prompts. Covers Task tool delegation, model selection, memory limits,\n  and declarative instruction design.\n\n  Use when: creating custom agents, designing agent descriptions for auto-delegation,\n  troubleshooting agent memory issues, or building agent pipelines.\nlicense: MIT\n---\n\n# Agent Development for Claude Code\n\nBuild effective custom agents for Claude Code with proper delegation, tool access, and prompt design.\n\n## Agent Description Pattern\n\nThe description field determines whether Claude will automatically delegate tasks.\n\n### Strong Trigger Pattern\n\n```yaml\n---\nname: agent-name\ndescription: |\n  [Role] specialist. MUST BE USED when [specific triggers].\n  Use PROACTIVELY for [task category].\n  Keywords: [trigger words]\ntools: Read, Write, Edit, Glob, Grep, Bash\nmodel: sonnet\n---\n```\n\n### Weak vs Strong Descriptions\n\n| Weak (won't auto-delegate) | Strong (auto-delegates) |\n|---------------------------|-------------------------|\n| \"Analyzes screenshots for issues\" | \"Visual QA specialist. MUST BE USED when analyzing screenshots. Use PROACTIVELY for visual QA.\" |\n| \"Runs Playwright scripts\" | \"Playwright specialist. MUST BE USED when running Playwright scripts. Use PROACTIVELY for browser automation.\" |\n\n**Key phrases**:\n- \"MUST BE USED when...\"\n- \"Use PROACTIVELY for...\"\n- Include trigger keywords\n\n### Delegation Mechanisms\n\n1. **Explicit**: `Task tool subagent_type: \"agent-name\"` - always works\n2. **Automatic**: Claude matches task to agent description - requires strong phrasing\n\n**Session restart required** after creating\u002Fmodifying agents.\n\n## Tool Access Principle\n\n**If an agent doesn't need Bash, don't give it Bash.**\n\n| Agent needs to... | Give tools | Don't give |\n|-------------------|------------|------------|\n| Create files only | Read, Write, Edit, Glob, Grep | Bash |\n| Run scripts\u002FCLIs | Read, Write, Edit, Glob, Grep, Bash | — |\n| Read\u002Faudit only | Read, Glob, Grep | Write, Edit, Bash |\n\n**Why?** Models default to `cat > file \u003C\u003C 'EOF'` heredocs instead of Write tool. Each bash command requires approval, causing dozens of prompts per agent run.\n\n### Allowlist Pattern\n\nInstead of restricting Bash, allowlist safe commands in `.claude\u002Fsettings.json`:\n\n```json\n{\n  \"permissions\": {\n    \"allow\": [\n      \"Write\", \"Edit\", \"WebFetch(domain:*)\",\n      \"Bash(cd *)\", \"Bash(cp *)\", \"Bash(mkdir *)\", \"Bash(ls *)\",\n      \"Bash(cat *)\", \"Bash(head *)\", \"Bash(tail *)\", \"Bash(grep *)\",\n      \"Bash(diff *)\", \"Bash(mv *)\", \"Bash(touch *)\", \"Bash(file *)\"\n    ]\n  }\n}\n```\n\n## Model Selection (Quality First)\n\nDon't downgrade quality to work around issues - fix root causes instead.\n\n| Model | Use For |\n|-------|---------|\n| **Opus** | Creative work (page building, design, content) - quality matters |\n| **Sonnet** | Most agents - content, code, research (default) |\n| **Haiku** | Only script runners where quality doesn't matter |\n\n## Memory Limits\n\n### Root Cause Fix (REQUIRED)\n\nAdd to `~\u002F.bashrc` or `~\u002F.zshrc`:\n```bash\nexport NODE_OPTIONS=\"--max-old-space-size=16384\"\n```\n\nIncreases Node.js heap from 4GB to 16GB.\n\n### Parallel Limits (Even With Fix)\n\n| Agent Type | Max Parallel | Notes |\n|------------|--------------|-------|\n| Any agents | 2-3 | Context accumulates; batch then pause |\n| Heavy creative (Opus) | 1-2 | Uses more memory |\n\n### Recovery\n\n1. `source ~\u002F.bashrc` or restart terminal\n2. `NODE_OPTIONS=\"--max-old-space-size=16384\" claude`\n3. Check what files exist, continue from there\n\n## Sub-Agent vs Remote API\n\n**Always prefer Task sub-agents over remote API calls.**\n\n| Aspect | Remote API Call | Task Sub-Agent |\n|--------|-----------------|----------------|\n| Tool access | None | Full (Read, Grep, Write, Bash) |\n| File reading | Must pass all content in prompt | Can read files iteratively |\n| Cross-referencing | Single context window | Can reason across documents |\n| Decision quality | Generic suggestions | Specific decisions with rationale |\n| Output quality | ~100 lines typical | 600+ lines with specifics |\n\n```typescript\n\u002F\u002F ❌ WRONG - Remote API call\nconst response = await fetch('https:\u002F\u002Fapi.anthropic.com\u002Fv1\u002Fmessages', {...})\n\n\u002F\u002F ✅ CORRECT - Use Task tool\n\u002F\u002F Invoke Task with subagent_type: \"general-purpose\"\n```\n\n## Declarative Over Imperative\n\nDescribe **what** to accomplish, not **how** to use tools.\n\n### Wrong (Imperative)\n\n```markdown\n### Check for placeholders\n```bash\ngrep -r \"PLACEHOLDER:\" build\u002F*.html\n```\n```\n\n### Right (Declarative)\n\n```markdown\n### Check for placeholders\nSearch all HTML files in build\u002F for:\n- PLACEHOLDER: comments\n- TODO or TBD markers\n- Template brackets like [Client Name]\n\nAny match = incomplete content.\n```\n\n### What to Include\n\n| Include | Skip |\n|---------|------|\n| Task goal and context | Explicit bash\u002Ftool commands |\n| Input file paths | \"Use X tool to...\" |\n| Output file paths and format | Step-by-step tool invocations |\n| Success\u002Ffailure criteria | Shell pipeline syntax |\n| Blocking checks (prerequisites) | Micromanaged workflows |\n| Quality checklists | |\n\n## Self-Documentation Principle\n\n> \"Agents that won't have your context must be able to reproduce the behaviour independently.\"\n\nEvery improvement must be encoded into the agent's prompt, not left as implicit knowledge.\n\n### What to Encode\n\n| Discovery | Where to Capture |\n|-----------|------------------|\n| Bug fix pattern | Agent's \"Corrections\" or \"Common Issues\" section |\n| Quality requirement | Agent's \"Quality Checklist\" section |\n| File path convention | Agent's \"Output\" section |\n| Tool usage pattern | Agent's \"Process\" section |\n| Blocking prerequisite | Agent's \"Blocking Check\" section |\n\n### Test: Would a Fresh Agent Succeed?\n\nBefore completing any agent improvement:\n1. Read the agent prompt as if you have no context\n2. Ask: Could a new session follow this and produce the same quality?\n3. If no: Add missing instructions, patterns, or references\n\n### Anti-Patterns\n\n| Anti-Pattern | Why It Fails |\n|--------------|--------------|\n| \"As we discussed earlier...\" | No prior context exists |\n| Relying on files read during dev | Agent may not read same files |\n| Assuming knowledge from errors | Agent won't see your debugging |\n| \"Just like the home page\" | Agent hasn't built home page |\n\n## Agent Prompt Structure\n\nEffective agent prompts include:\n\n```markdown\n## Your Role\n[What the agent does]\n\n## Blocking Check\n[Prerequisites that must exist]\n\n## Input\n[What files to read]\n\n## Process\n[Step-by-step with encoded learnings]\n\n## Output\n[Exact file paths and formats]\n\n## Quality Checklist\n[Verification steps including learned gotchas]\n\n## Common Issues\n[Patterns discovered during development]\n```\n\n## Pipeline Agents\n\nWhen inserting a new agent into a numbered pipeline (e.g., `HTML-01` → `HTML-05` → `HTML-11`):\n\n| Must Update | What |\n|-------------|------|\n| New agent | \"Workflow Position\" diagram + \"Next\" field |\n| **Predecessor agent** | Its \"Next\" field to point to new agent |\n\n**Common bug**: New agent is \"orphaned\" because predecessor still points to old next agent.\n\n**Verification**:\n```bash\ngrep -n \"Next:.*→\\|Then.*runs next\" .claude\u002Fagents\u002F*.md\n```\n\n## The Sweet Spot\n\n**Best use case**: Tasks that are **repetitive but require judgment**.\n\nExample: Auditing 70 skills manually = tedious. But each audit needs intelligence (check docs, compare versions, decide what to fix). Perfect for parallel agents with clear instructions.\n\n**Not good for**:\n- Simple tasks (just do them)\n- Highly creative tasks (need human direction)\n- Tasks requiring cross-file coordination (agents work independently)\n\n## Effective Prompt Template\n\n```\nFor each [item]:\n1. Read [source file]\n2. Verify with [external check - npm view, API call, etc.]\n3. Check [authoritative source]\n4. Score\u002Fevaluate\n5. FIX issues found ← Critical instruction\n```\n\n**Key elements**:\n- **\"FIX issues found\"** - Without this, agents only report. With it, they take action.\n- **Exact file paths** - Prevents ambiguity\n- **Output format template** - Ensures consistent, parseable reports\n- **Batch size ~5 items** - Enough work to be efficient, not so much that failures cascade\n\n## Workflow Pattern\n\n```\n1. ME: Launch 2-3 parallel agents with identical prompt, different item lists\n2. AGENTS: Work in parallel (read → verify → check → edit → report)\n3. AGENTS: Return structured reports (score, status, fixes applied, files modified)\n4. ME: Review changes (git status, spot-check diffs)\n5. ME: Commit in batches with meaningful changelog\n6. ME: Push and update progress tracking\n```\n\n**Why agents don't commit**: Allows human review, batching, and clean commit history.\n\n## Signs a Task Fits This Pattern\n\n**Good fit**:\n- Same steps repeated for many items\n- Each item requires judgment (not just transformation)\n- Items are independent (no cross-item dependencies)\n- Clear success criteria (score, pass\u002Ffail, etc.)\n- Authoritative source exists to verify against\n\n**Bad fit**:\n- Items depend on each other's results\n- Requires creative\u002Fsubjective decisions\n- Single complex task (use regular agent instead)\n- Needs human input mid-process\n\n## Quick Reference\n\n### Agent Frontmatter Template\n\n```yaml\n---\nname: my-agent\ndescription: |\n  [Role] specialist. MUST BE USED when [triggers].\n  Use PROACTIVELY for [task category].\n  Keywords: [trigger words]\ntools: Read, Write, Edit, Glob, Grep, Bash\nmodel: sonnet\n---\n```\n\n### Fix Bash Approval Spam\n\n1. Remove Bash from tools if not needed\n2. Put critical instructions FIRST (right after frontmatter)\n3. Use allowlists in `.claude\u002Fsettings.json`\n\n### Memory Crash Recovery\n\n```bash\nexport NODE_OPTIONS=\"--max-old-space-size=16384\"\nsource ~\u002F.bashrc && claude\n```\n","# Agent Development for Claude Code\n\nBuild effective custom agents for Claude Code with proper delegation, tool access, and prompt design.\n\n## Agent Description Pattern\n\nThe description field determines whether Claude will automatically delegate tasks.\n\n### Strong Trigger Pattern\n\n```yaml\n---\nname: agent-name\ndescription: |\n  [Role] specialist. MUST BE USED when [specific triggers].\n  Use PROACTIVELY for [task category].\n  Keywords: [trigger words]\ntools: Read, Write, Edit, Glob, Grep, Bash\nmodel: sonnet\n---\n```\n\n### Weak vs Strong Descriptions\n\n| Weak (won't auto-delegate) | Strong (auto-delegates) |\n|---------------------------|-------------------------|\n| \"Analyzes screenshots for issues\" | \"Visual QA specialist. MUST BE USED when analyzing screenshots. Use PROACTIVELY for visual QA.\" |\n| \"Runs Playwright scripts\" | \"Playwright specialist. MUST BE USED when running Playwright scripts. Use PROACTIVELY for browser automation.\" |\n\n**Key phrases**:\n- \"MUST BE USED when...\"\n- \"Use PROACTIVELY for...\"\n- Include trigger keywords\n\n### Delegation Mechanisms\n\n1. **Explicit**: `Task tool subagent_type: \"agent-name\"` - always works\n2. **Automatic**: Claude matches task to agent description - requires strong phrasing\n\n**Session restart required** after creating\u002Fmodifying agents.\n\n## Tool Access Principle\n\n**If an agent doesn't need Bash, don't give it Bash.**\n\n| Agent needs to... | Give tools | Don't give |\n|-------------------|------------|------------|\n| Create files only | Read, Write, Edit, Glob, Grep | Bash |\n| Run scripts\u002FCLIs | Read, Write, Edit, Glob, Grep, Bash | — |\n| Read\u002Faudit only | Read, Glob, Grep | Write, Edit, Bash |\n\n**Why?** Models default to `cat > file \u003C\u003C 'EOF'` heredocs instead of Write tool. Each bash command requires approval, causing dozens of prompts per agent run.\n\n### Allowlist Pattern\n\nInstead of restricting Bash, allowlist safe commands in `.claude\u002Fsettings.json`:\n\n```json\n{\n  \"permissions\": {\n    \"allow\": [\n      \"Write\", \"Edit\", \"WebFetch(domain:*)\",\n      \"Bash(cd *)\", \"Bash(cp *)\", \"Bash(mkdir *)\", \"Bash(ls *)\",\n      \"Bash(cat *)\", \"Bash(head *)\", \"Bash(tail *)\", \"Bash(grep *)\",\n      \"Bash(diff *)\", \"Bash(mv *)\", \"Bash(touch *)\", \"Bash(file *)\"\n    ]\n  }\n}\n```\n\n## Model Selection (Quality First)\n\nDon't downgrade quality to work around issues - fix root causes instead.\n\n| Model | Use For |\n|-------|---------|\n| **Opus** | Creative work (page building, design, content) - quality matters |\n| **Sonnet** | Most agents - content, code, research (default) |\n| **Haiku** | Only script runners where quality doesn't matter |\n\n## Memory Limits\n\n### Root Cause Fix (REQUIRED)\n\nAdd to `~\u002F.bashrc` or `~\u002F.zshrc`:\n```bash\nexport NODE_OPTIONS=\"--max-old-space-size=16384\"\n```\n\nIncreases Node.js heap from 4GB to 16GB.\n\n### Parallel Limits (Even With Fix)\n\n| Agent Type | Max Parallel | Notes |\n|------------|--------------|-------|\n| Any agents | 2-3 | Context accumulates; batch then pause |\n| Heavy creative (Opus) | 1-2 | Uses more memory |\n\n### Recovery\n\n1. `source ~\u002F.bashrc` or restart terminal\n2. `NODE_OPTIONS=\"--max-old-space-size=16384\" claude`\n3. Check what files exist, continue from there\n\n## Sub-Agent vs Remote API\n\n**Always prefer Task sub-agents over remote API calls.**\n\n| Aspect | Remote API Call | Task Sub-Agent |\n|--------|-----------------|----------------|\n| Tool access | None | Full (Read, Grep, Write, Bash) |\n| File reading | Must pass all content in prompt | Can read files iteratively |\n| Cross-referencing | Single context window | Can reason across documents |\n| Decision quality | Generic suggestions | Specific decisions with rationale |\n| Output quality | ~100 lines typical | 600+ lines with specifics |\n\n```typescript\n\u002F\u002F ❌ WRONG - Remote API call\nconst response = await fetch('https:\u002F\u002Fapi.anthropic.com\u002Fv1\u002Fmessages', {...})\n\n\u002F\u002F ✅ CORRECT - Use Task tool\n\u002F\u002F Invoke Task with subagent_type: \"general-purpose\"\n```\n\n## Declarative Over Imperative\n\nDescribe **what** to accomplish, not **how** to use tools.\n\n### Wrong (Imperative)\n\n```markdown\n### Check for placeholders\n```bash\ngrep -r \"PLACEHOLDER:\" build\u002F*.html\n```\n```\n\n### Right (Declarative)\n\n```markdown\n### Check for placeholders\nSearch all HTML files in build\u002F for:\n- PLACEHOLDER: comments\n- TODO or TBD markers\n- Template brackets like [Client Name]\n\nAny match = incomplete content.\n```\n\n### What to Include\n\n| Include | Skip |\n|---------|------|\n| Task goal and context | Explicit bash\u002Ftool commands |\n| Input file paths | \"Use X tool to...\" |\n| Output file paths and format | Step-by-step tool invocations |\n| Success\u002Ffailure criteria | Shell pipeline syntax |\n| Blocking checks (prerequisites) | Micromanaged workflows |\n| Quality checklists | |\n\n## Self-Documentation Principle\n\n> \"Agents that won't have your context must be able to reproduce the behaviour independently.\"\n\nEvery improvement must be encoded into the agent's prompt, not left as implicit knowledge.\n\n### What to Encode\n\n| Discovery | Where to Capture |\n|-----------|------------------|\n| Bug fix pattern | Agent's \"Corrections\" or \"Common Issues\" section |\n| Quality requirement | Agent's \"Quality Checklist\" section |\n| File path convention | Agent's \"Output\" section |\n| Tool usage pattern | Agent's \"Process\" section |\n| Blocking prerequisite | Agent's \"Blocking Check\" section |\n\n### Test: Would a Fresh Agent Succeed?\n\nBefore completing any agent improvement:\n1. Read the agent prompt as if you have no context\n2. Ask: Could a new session follow this and produce the same quality?\n3. If no: Add missing instructions, patterns, or references\n\n### Anti-Patterns\n\n| Anti-Pattern | Why It Fails |\n|--------------|--------------|\n| \"As we discussed earlier...\" | No prior context exists |\n| Relying on files read during dev | Agent may not read same files |\n| Assuming knowledge from errors | Agent won't see your debugging |\n| \"Just like the home page\" | Agent hasn't built home page |\n\n## Agent Prompt Structure\n\nEffective agent prompts include:\n\n```markdown\n## Your Role\n[What the agent does]\n\n## Blocking Check\n[Prerequisites that must exist]\n\n## Input\n[What files to read]\n\n## Process\n[Step-by-step with encoded learnings]\n\n## Output\n[Exact file paths and formats]\n\n## Quality Checklist\n[Verification steps including learned gotchas]\n\n## Common Issues\n[Patterns discovered during development]\n```\n\n## Pipeline Agents\n\nWhen inserting a new agent into a numbered pipeline (e.g., `HTML-01` → `HTML-05` → `HTML-11`):\n\n| Must Update | What |\n|-------------|------|\n| New agent | \"Workflow Position\" diagram + \"Next\" field |\n| **Predecessor agent** | Its \"Next\" field to point to new agent |\n\n**Common bug**: New agent is \"orphaned\" because predecessor still points to old next agent.\n\n**Verification**:\n```bash\ngrep -n \"Next:.*→\\|Then.*runs next\" .claude\u002Fagents\u002F*.md\n```\n\n## The Sweet Spot\n\n**Best use case**: Tasks that are **repetitive but require judgment**.\n\nExample: Auditing 70 skills manually = tedious. But each audit needs intelligence (check docs, compare versions, decide what to fix). Perfect for parallel agents with clear instructions.\n\n**Not good for**:\n- Simple tasks (just do them)\n- Highly creative tasks (need human direction)\n- Tasks requiring cross-file coordination (agents work independently)\n\n## Effective Prompt Template\n\n```\nFor each [item]:\n1. Read [source file]\n2. Verify with [external check - npm view, API call, etc.]\n3. Check [authoritative source]\n4. Score\u002Fevaluate\n5. FIX issues found ← Critical instruction\n```\n\n**Key elements**:\n- **\"FIX issues found\"** - Without this, agents only report. With it, they take action.\n- **Exact file paths** - Prevents ambiguity\n- **Output format template** - Ensures consistent, parseable reports\n- **Batch size ~5 items** - Enough work to be efficient, not so much that failures cascade\n\n## Workflow Pattern\n\n```\n1. ME: Launch 2-3 parallel agents with identical prompt, different item lists\n2. AGENTS: Work in parallel (read → verify → check → edit → report)\n3. AGENTS: Return structured reports (score, status, fixes applied, files modified)\n4. ME: Review changes (git status, spot-check diffs)\n5. ME: Commit in batches with meaningful changelog\n6. ME: Push and update progress tracking\n```\n\n**Why agents don't commit**: Allows human review, batching, and clean commit history.\n\n## Signs a Task Fits This Pattern\n\n**Good fit**:\n- Same steps repeated for many items\n- Each item requires judgment (not just transformation)\n- Items are independent (no cross-item dependencies)\n- Clear success criteria (score, pass\u002Ffail, etc.)\n- Authoritative source exists to verify against\n\n**Bad fit**:\n- Items depend on each other's results\n- Requires creative\u002Fsubjective decisions\n- Single complex task (use regular agent instead)\n- Needs human input mid-process\n\n## Quick Reference\n\n### Agent Frontmatter Template\n\n```yaml\n---\nname: my-agent\ndescription: |\n  [Role] specialist. MUST BE USED when [triggers].\n  Use PROACTIVELY for [task category].\n  Keywords: [trigger words]\ntools: Read, Write, Edit, Glob, Grep, Bash\nmodel: sonnet\n---\n```\n\n### Fix Bash Approval Spam\n\n1. Remove Bash from tools if not needed\n2. Put critical instructions FIRST (right after frontmatter)\n3. Use allowlists in `.claude\u002Fsettings.json`\n\n### Memory Crash Recovery\n\n```bash\nexport NODE_OPTIONS=\"--max-old-space-size=16384\"\nsource ~\u002F.bashrc && claude\n```\n",{"name":5,"license":15,"description":54},"Design and build custom Claude Code agents with effective descriptions, tool access patterns,\nand self-documenting prompts. Covers Task tool delegation, model selection, memory limits,\nand declarative instruction design.\n\nUse when: creating custom agents, designing agent descriptions for auto-delegation,\ntroubleshooting agent memory issues, or building agent pipelines.\n","main","b80cde2ef85241dd4fd54a1b934f5e22835f78ba","skills\u002Fagent-development",102,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills","https:\u002F\u002Fgithub.com\u002Fsundial-org","https:\u002F\u002Fwww.sundialhub.com","2026-03-07 14:57:19",{"source":8,"skillName":5,"commands":64,"prompt":74,"zipAvailable":75},[65,68,71],{"tool":66,"command":67},"npx","npx skills add sundial-org\u002Fawesome-openclaw-skills --skill agent-development",{"tool":69,"command":70},"bunx","bunx skills add sundial-org\u002Fawesome-openclaw-skills --skill agent-development",{"tool":72,"command":73},"pnpm","pnpm dlx skills add sundial-org\u002Fawesome-openclaw-skills --skill agent-development","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002FSKILL.md",true,[77,83,87,91,95,99,103],{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},".claude-plugin\u002Fplugin.json","REFERENCE","参考文档",735,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002F.claude-plugin\u002Fplugin.json",{"path":84,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":85,"sourceHtmlUrl":86},"README.md",2198,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002FREADME.md",{"path":88,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":89,"sourceHtmlUrl":90},"rules\u002Fagent-memory-limits.md",1024,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002Frules\u002Fagent-memory-limits.md",{"path":92,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":93,"sourceHtmlUrl":94},"rules\u002Fagent-pattern.md",2291,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002Frules\u002Fagent-pattern.md",{"path":96,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":97,"sourceHtmlUrl":98},"rules\u002Fagent-self-documentation.md",2953,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002Frules\u002Fagent-self-documentation.md",{"path":100,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":101,"sourceHtmlUrl":102},"rules\u002Fcustom-agent-descriptions.md",2133,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002Frules\u002Fcustom-agent-descriptions.md",{"path":104,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":105,"sourceHtmlUrl":106},"rules\u002Fcustom-agent-instructions.md",3738,"https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fagent-development\u002Frules\u002Fcustom-agent-instructions.md",[108,118,146,168,192,216],{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":49},"01M6CGQVNAEMJTA5P9GZVYNCVG","x-trends","Fetches current top trending topics on X (Twitter) for any country using public aggregators.","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fx-trends\u002FSKILL.md",[],[115,116,117],"x trends","public","twitter",{"skillNo":119,"name":120,"nameZh":6,"description":121,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":122,"categories":123,"keywords":127,"publishedAt":49},"01M6CGKR0DGWD64YAYJ4WH32C0","vercel-react-best-practices","React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React\u002FNext.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-react-best-practices\u002FSKILL.md",[124],{"categoryNo":125,"categoryName":126,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000009","前端开发",[128,129,130,131,132,133,134,135,136,137,138,139,140,46,141,142,143,144,145],"vercel react best practices","react","best practices","react best practices","vercel","tasks","performance optimization","js","refactoring","Next.js","data","Optimization","engineering","guidelines","react components","components","performance","writing",{"skillNo":147,"name":148,"nameZh":6,"description":149,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":150,"categories":151,"keywords":152,"publishedAt":49},"01M6CGGNS5N14QNV719AK25G94","topydo","Manage todo.txt tasks using topydo CLI. Add, list, complete, prioritize, tag, and organize tasks with dependencies, due dates, recurrence, and projects. Use for any task management, todo lists, or when the user mentions tasks, todos, or todo.txt. Requires Python 3 and pip. Works on macOS, Linux, and Windows.","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Ftopydo\u002FSKILL.md",[],[148,153,133,154,155,156,157,158,159,160,161,162,163,164,165,166,36,167],"windows","prioritize","linux","projects","task management","list","github","python","cli","macos","management","todo","task","manage","user",{"skillNo":169,"name":170,"nameZh":6,"description":171,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":172,"categories":173,"keywords":174,"publishedAt":49},"01M6CFRM7YJ2Q07KTV6CH0PMCK","openclaw-feeds","RSS news aggregator. Fetches headlines from curated feeds across three categories: news, games, and finance. Use when the user asks about current news, headlines, what's happening, what's going on, or says \"what's up in news\", \"what's up in finance\", \"what's up in games\", or the German equivalents \"was geht mit nachrichten\", \"was geht mit money\", \"was geht mit gaming\". Also activates for requests like \"give me a news rundown\", \"latest headlines\", \"market news\", \"gaming news\", \"tech news\", \"finance roundup\", or \"briefing\". Returns structured JSON from public RSS feeds — no API keys, no web search needed.","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fopenclaw-feeds\u002FSKILL.md",[],[175,176,177,178,179,180,181,36,182,183,184,185,40,186,187,188,189,190,160,191],"openclaw","feeds","openclaw feeds","fetch","german","Market","up","install","json","rss","web search","Bash","api","roundup","search","web","python3",{"skillNo":193,"name":194,"nameZh":6,"description":195,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":196,"categories":197,"keywords":202,"publishedAt":49},"01M6CFQMN6XK3X0D0FFAJ4ZCC0","octolens","Query and analyze brand mentions from Octolens API. Use when the user wants to fetch mentions, track keywords, filter by source platforms (Twitter, Reddit, GitHub, LinkedIn, etc.), sentiment analysis, or analyze social media engagement. Supports complex filtering with AND\u002FOR logic, date ranges, follower counts, and bookmarks.","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Foctolens\u002FSKILL.md",[198,201],{"categoryNo":199,"categoryName":200,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},[194,203,187,204,205,206,159,40,36,207,208,178,209,210,211,212,117,213,214,215],"Media","analysis","sentiment","sentiment analysis","engagement","filtering","read","bookmarks","social media","linkedin","node","brand","analyze",{"skillNo":217,"name":218,"nameZh":6,"description":219,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":220,"categories":221,"keywords":223,"publishedAt":49},"01M6CFEG2QX0E4Y9D5VFXWGQ2X","lmstudio-subagents","Reduces token usage from paid providers by offloading work to local LM Studio models. Use when: (1) Cutting costs—use local models for summarization, extraction, classification, rewriting, first-pass review, brainstorming when quality suffices, (2) Avoiding paid API calls for high-volume or repetitive tasks, (3) No extra model configuration—JIT loading and REST API work with existing LM Studio setup, (4) Local-only or privacy-sensitive work. Requires LM Studio 0.4+ with server (default :1234). No CLI required.","https:\u002F\u002Fgithub.com\u002Fsundial-org\u002Fawesome-openclaw-skills\u002Fblob\u002Fmain\u002Fskills\u002Flmstudio-subagents\u002FSKILL.md",[222],{"categoryNo":199,"categoryName":200,"taxonomyType":21,"taxonomyTypeDesc":22},[224,225,226,227,228,229,36,230,37,231,187,232,233,234,235,236,133,237,161,238],"lmstudio subagents","subagents","quality","default","models","only","work","classification","first","REST","Local Models","privacy","server","brainstorming","review"]