[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-microsoft-role-definitions":3,"mdc-53z5ec-key":34,"related-org-microsoft-role-definitions":1416,"related-repo-microsoft-role-definitions":1609},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"role-definitions","configure model routing and fallback chains","Detailed model routing role definitions, decision flowchart, model tier grid, and fallback chain best practices. Load when authoring agents or debugging model routing.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},"microsoft","Microsoft","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fmicrosoft.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"LLM","llm","tag",{"name":17,"slug":18,"type":15},"Architecture","architecture",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Agents","agents",2,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Famplifier-bundle-routing-matrix","2026-07-24T05:42:25.781133",null,10,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"Routing Matrix bundle for the Amplifier project","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Famplifier-bundle-routing-matrix\u002Ftree\u002FHEAD\u002Fskills\u002Frole-definitions","---\nname: role-definitions\ndescription: 'Detailed model routing role definitions, decision flowchart, model tier grid, and fallback chain best practices. Load when authoring agents or debugging model routing.'\nversion: 1.0.0\n---\n\n# Model Routing Role Definitions\n\nReference for choosing the right `model_role` when writing agent frontmatter or delegating to sub-agents. There are 13 roles organized into 5 categories.\n\n## Quick Decision Flowchart\n\n```\nWhat does your agent primarily DO?\n│\n├─ Writes code?\n│   ├─ Frontend\u002FUI code (components, layouts, CSS)? → ui-coding\n│   ├─ Security vulnerability scanning? → security-audit\n│   └─ General code (backend, IaC, tests, debugging)? → coding\n│\n├─ Thinks deeply \u002F designs systems?\n│   ├─ Designing architecture or solving complex problems? → reasoning\n│   ├─ Reviewing\u002Fcritiquing existing work for flaws? → critique\n│   └─ Investigating and synthesizing from multiple sources? → research\n│\n├─ Creates content?\n│   ├─ Aesthetic\u002Fdesign direction (visual, brand, style)? → creative\n│   └─ Long-form writing (docs, marketing, case studies)? → writing\n│\n├─ Works with images?\n│   ├─ Needs to understand\u002Fanalyze images? → vision\n│   └─ Needs to generate images? → image-gen\n│\n├─ Orchestrates high-stakes operations?\n│   └─ Infrastructure, deployments, shadow environments? → critical-ops\n│\n├─ Quick utility task (parsing, classification, file ops)?\n│   └─ → fast\n│\n└─ None of the above \u002F genuinely varied work?\n    └─ → general\n```\n\n## Role-by-Role Reference\n\n### Foundation Roles\n\nThese two roles are required in every matrix and serve as universal fallbacks.\n\n#### `general`\n- **Description:** Versatile catch-all, no specialization needed\n- **Model tier:** Mid (Sonnet, gpt-5.5, Gemini Pro)\n- **When to use:** The agent's work is genuinely varied — knowledge experts, ecosystem consultants, integration specialists\n- **When NOT to use:** If you can name what the agent primarily does (codes, designs, reviews), use that role instead\n- **Example agents:** explorer, foundation-expert, amplifier-expert, core-expert\n- **Example chains:** `general` (terminal — no fallback needed)\n\n#### `fast`\n- **Description:** Quick utility tasks — parsing, classification, file ops, bulk work\n- **Model tier:** Cheap (Haiku, GPT luna, Gemini Flash)\n- **When to use:** Well-defined, low-ambiguity, high-volume, or latency-sensitive tasks — file operations, git commands, notification triage, data extraction\n- **When NOT to use:** If the agent needs judgment calls, synthesis, or quality prose\n- **Example agents:** file-ops, git-ops, shell-exec, health-checker, triage-manager\n- **Example chains:** `fast` (terminal — fast agents rarely need fallback)\n\n---\n\n### Coding Domain Roles\n\nFor agents whose primary output is code.\n\n#### `coding`\n- **Description:** Code generation, implementation, debugging\n- **Model tier:** Mid, code-specialized (Sonnet, gpt-5.5, Gemini Pro)\n- **When to use:** The agent writes, modifies, or debugs code as its primary activity — bug fixing, feature implementation, test writing, infrastructure-as-code\n- **When NOT to use:** If the agent primarily reviews code (use `critique`), designs UI layouts (use `ui-coding`), or audits for vulnerabilities (use `security-audit`)\n- **Example agents:** bug-hunter, modular-builder, test-coverage, python-dev, rust-dev\n- **Example chains:** `[coding, general]`\n\n#### `ui-coding`\n- **Description:** Frontend\u002FUI code — components, layouts, styling, spatial reasoning\n- **Model tier:** Mid, code-specialized (same as `coding` today — will diverge as visually-tuned models emerge)\n- **When to use:** The agent builds user-facing interfaces — components, layouts, responsive design, CSS, accessibility\n- **When NOT to use:** If the agent writes backend code that serves a frontend (use `coding`). If it evaluates design aesthetics without writing code (use `creative`)\n- **Example agents:** component-designer, layout-architect, responsive-strategist\n- **Example chains:** `[ui-coding, coding, general]`\n\n#### `security-audit`\n- **Description:** Vulnerability assessment, attack surface analysis, code auditing\n- **Model tier:** Mid, code-specialized + high reasoning\n- **When to use:** The agent examines code or systems for security vulnerabilities, reviews authentication flows, assesses attack surfaces\n- **When NOT to use:** If the agent is doing general code review (use `critique`). If it's implementing security features (use `coding`)\n- **Example agents:** security-guardian\n- **Example chains:** `[security-audit, critique, general]`\n\n---\n\n### Cognitive Mode Roles\n\nFor agents whose primary value is in HOW they think.\n\n#### `reasoning`\n- **Description:** Deep architectural reasoning, system design, complex multi-step analysis\n- **Model tier:** Heavy (Opus, gpt-5.5-pro, Gemini Pro) with high reasoning effort\n- **When to use:** The agent designs systems, plans architectures, breaks complex problems into steps, or needs extended chain-of-thought\n- **When NOT to use:** If the agent evaluates existing work (use `critique`), writes prose (use `writing`), or generates creative concepts (use `creative`)\n- **Example agents:** zen-architect, brainstormer, plan-writer, recipe-author\n- **Example chains:** `[reasoning, general]`\n\n#### `critique`\n- **Description:** Analytical evaluation — finding flaws in existing work, not generating solutions\n- **Model tier:** Mid with extra-high reasoning effort\n- **When to use:** The agent reviews, evaluates, or finds flaws in existing code, architecture, or plans\n- **When NOT to use:** If the agent builds something new (use `reasoning` or `coding`). If it checks for security vulnerabilities specifically (use `security-audit`)\n- **Example agents:** spec-reviewer, code-quality-reviewer, friction-detector\n- **Example chains:** `[critique, reasoning, general]` or `[critique, general]`\n\n#### `creative`\n- **Description:** Design direction, aesthetic judgment, high-quality creative output\n- **Model tier:** Heavy (Opus, gpt-5.5, Gemini Pro)\n- **When to use:** The agent makes aesthetic judgments, establishes design direction, creates visual concepts\n- **When NOT to use:** If the agent writes long-form content (use `writing`). If it writes UI code (use `ui-coding`)\n- **Example agents:** art-director, style-curator, storyboard-writer, character-designer\n- **Example chains:** `[creative, general]` or `[creative, writing, general]`\n\n#### `writing`\n- **Description:** Long-form content — documentation, marketing, case studies, storytelling\n- **Model tier:** Heavy (Opus, gpt-5.5, Gemini Pro)\n- **When to use:** The agent produces sustained written content — documentation, blog posts, case studies, release notes\n- **When NOT to use:** If the output is code with comments (use `coding`). If the writing is short utility text (use `fast`)\n- **Example agents:** technical-writer, marketing-writer, release-manager\n- **Example chains:** `[writing, creative, general]` or `[writing, general]`\n\n#### `research`\n- **Description:** Deep investigation, information synthesis across multiple sources\n- **Model tier:** Heavy (Opus, gpt-5.5-pro, Gemini Pro) with high reasoning effort\n- **When to use:** The agent investigates and synthesizes from multiple sources. Extended context windows matter\n- **When NOT to use:** If the agent designs based on what it already knows (use `reasoning`)\n- **Example agents:** browser-researcher, story-researcher\n- **Example chains:** `[research, general]`\n\n---\n\n### Capability Roles\n\nFor agents that need specific model capabilities beyond text.\n\n#### `vision`\n- **Description:** Understanding visual input — screenshots, diagrams, UI mockups\n- **Model tier:** Mid, multimodal (Gemini Flash, Sonnet, gpt-5.5)\n- **When to use:** The agent analyzes screenshots, reads diagrams, interprets UI mockups\n- **When NOT to use:** If the agent generates images (use `image-gen`). If it writes UI code without seeing existing UI (use `ui-coding`)\n- **Example agents:** browser-operator, visual-documenter\n- **Example chains:** `[vision, general]`\n\n#### `image-gen`\n- **Description:** Image generation, visual mockup creation, visual ideation\n- **Model tier:** Specialized (gemini-3-pro-image-preview) — sparse provider coverage\n- **When to use:** The agent creates images — generating mockups, producing comic panels, visual prototypes\n- **When NOT to use:** If the agent analyzes existing images (use `vision`)\n- **IMPORTANT:** Always include a non-image-gen fallback. Google-only in most matrices today.\n- **Example agents:** panel-artist, cover-artist\n- **Example chains:** `[image-gen, creative, general]` (NEVER `image-gen` alone)\n\n---\n\n### Operational Role\n\n#### `critical-ops`\n- **Description:** High-reliability operational tasks — infrastructure, orchestration, coordination where mistakes are costly\n- **Model tier:** Heavy (Opus, gpt-5.5-pro, Gemini Pro)\n- **When to use:** The agent orchestrates infrastructure, manages deployments, verifies shadow environments, or performs operational tasks where failures cascade\n- **When NOT to use:** If the agent writes IaC as its primary activity (use `coding`). If it does quick operational checks (use `fast`)\n- **Example agents:** setup-digital-twin, dtu-profile-builder, container-operator\n- **Example chains:** `[critical-ops, coding, general]`\n\n---\n\n## Model Tier Grid\n\nEvery role maps to a unique (model-tier x reasoning-config) cell:\n\n```\n                default reasoning        high reasoning         extra-high reasoning\n             ┌──────────────────────┬──────────────────────┬──────────────────────┐\n  Heavy      │  creative            │  reasoning           │                      │\n  (Opus)     │  writing             │  research            │                      │\n             │  critical-ops        │                      │                      │\n             ├──────────────────────┼──────────────────────┼──────────────────────┤\n  Mid        │  general   coding    │  security-audit      │  critique            │\n  (Sonnet)   │  ui-coding           │                      │                      │\n             ├──────────────────────┤                      │                      │\n  Flash      │  fast     vision     │                      │                      │\n  (Haiku)    │                      │                      │                      │\n             ├──────────────────────┤                      │                      │\n  Specialized│  image-gen           │                      │                      │\n             └──────────────────────┴──────────────────────┴──────────────────────┘\n```\n\n## Fallback Chain Best Practices\n\n1. **Every chain should end with `general`** (or `fast` for utility agents)\n2. **Go specific → general:** `[ui-coding, coding, general]` not `[general, ui-coding]`\n3. **Sparse roles need fallbacks:** `[image-gen, creative, general]` — image-gen has limited provider coverage\n4. **Don't over-chain:** 2-3 roles is typical. More than 4 suggests confusion about the agent's purpose\n5. **Single role is fine** when the agent's need is unambiguous: `fast`, `general`, `coding`\n",{"data":35,"body":37},{"name":4,"description":6,"version":36},"1.0.0",{"type":38,"children":39},"root",[40,49,64,71,83,89,96,101,111,184,193,256,260,266,271,280,366,374,457,465,540,543,549,554,563,646,654,743,751,833,841,922,931,998,1001,1007,1012,1021,1097,1105,1192,1195,1201,1210,1285,1288,1294,1299,1308,1314],{"type":41,"tag":42,"props":43,"children":45},"element","h1",{"id":44},"model-routing-role-definitions",[46],{"type":47,"value":48},"text","Model Routing Role Definitions",{"type":41,"tag":50,"props":51,"children":52},"p",{},[53,55,62],{"type":47,"value":54},"Reference for choosing the right ",{"type":41,"tag":56,"props":57,"children":59},"code",{"className":58},[],[60],{"type":47,"value":61},"model_role",{"type":47,"value":63}," when writing agent frontmatter or delegating to sub-agents. There are 13 roles organized into 5 categories.",{"type":41,"tag":65,"props":66,"children":68},"h2",{"id":67},"quick-decision-flowchart",[69],{"type":47,"value":70},"Quick Decision Flowchart",{"type":41,"tag":72,"props":73,"children":77},"pre",{"className":74,"code":76,"language":47},[75],"language-text","What does your agent primarily DO?\n│\n├─ Writes code?\n│   ├─ Frontend\u002FUI code (components, layouts, CSS)? → ui-coding\n│   ├─ Security vulnerability scanning? → security-audit\n│   └─ General code (backend, IaC, tests, debugging)? → coding\n│\n├─ Thinks deeply \u002F designs systems?\n│   ├─ Designing architecture or solving complex problems? → reasoning\n│   ├─ Reviewing\u002Fcritiquing existing work for flaws? → critique\n│   └─ Investigating and synthesizing from multiple sources? → research\n│\n├─ Creates content?\n│   ├─ Aesthetic\u002Fdesign direction (visual, brand, style)? → creative\n│   └─ Long-form writing (docs, marketing, case studies)? → writing\n│\n├─ Works with images?\n│   ├─ Needs to understand\u002Fanalyze images? → vision\n│   └─ Needs to generate images? → image-gen\n│\n├─ Orchestrates high-stakes operations?\n│   └─ Infrastructure, deployments, shadow environments? → critical-ops\n│\n├─ Quick utility task (parsing, classification, file ops)?\n│   └─ → fast\n│\n└─ None of the above \u002F genuinely varied work?\n    └─ → general\n",[78],{"type":41,"tag":56,"props":79,"children":81},{"__ignoreMap":80},"",[82],{"type":47,"value":76},{"type":41,"tag":65,"props":84,"children":86},{"id":85},"role-by-role-reference",[87],{"type":47,"value":88},"Role-by-Role Reference",{"type":41,"tag":90,"props":91,"children":93},"h3",{"id":92},"foundation-roles",[94],{"type":47,"value":95},"Foundation Roles",{"type":41,"tag":50,"props":97,"children":98},{},[99],{"type":47,"value":100},"These two roles are required in every matrix and serve as universal fallbacks.",{"type":41,"tag":102,"props":103,"children":105},"h4",{"id":104},"general",[106],{"type":41,"tag":56,"props":107,"children":109},{"className":108},[],[110],{"type":47,"value":104},{"type":41,"tag":112,"props":113,"children":114},"ul",{},[115,127,137,147,157,167],{"type":41,"tag":116,"props":117,"children":118},"li",{},[119,125],{"type":41,"tag":120,"props":121,"children":122},"strong",{},[123],{"type":47,"value":124},"Description:",{"type":47,"value":126}," Versatile catch-all, no specialization needed",{"type":41,"tag":116,"props":128,"children":129},{},[130,135],{"type":41,"tag":120,"props":131,"children":132},{},[133],{"type":47,"value":134},"Model tier:",{"type":47,"value":136}," Mid (Sonnet, gpt-5.5, Gemini Pro)",{"type":41,"tag":116,"props":138,"children":139},{},[140,145],{"type":41,"tag":120,"props":141,"children":142},{},[143],{"type":47,"value":144},"When to use:",{"type":47,"value":146}," The agent's work is genuinely varied — knowledge experts, ecosystem consultants, integration specialists",{"type":41,"tag":116,"props":148,"children":149},{},[150,155],{"type":41,"tag":120,"props":151,"children":152},{},[153],{"type":47,"value":154},"When NOT to use:",{"type":47,"value":156}," If you can name what the agent primarily does (codes, designs, reviews), use that role instead",{"type":41,"tag":116,"props":158,"children":159},{},[160,165],{"type":41,"tag":120,"props":161,"children":162},{},[163],{"type":47,"value":164},"Example agents:",{"type":47,"value":166}," explorer, foundation-expert, amplifier-expert, core-expert",{"type":41,"tag":116,"props":168,"children":169},{},[170,175,177,182],{"type":41,"tag":120,"props":171,"children":172},{},[173],{"type":47,"value":174},"Example chains:",{"type":47,"value":176}," ",{"type":41,"tag":56,"props":178,"children":180},{"className":179},[],[181],{"type":47,"value":104},{"type":47,"value":183}," (terminal — no fallback needed)",{"type":41,"tag":102,"props":185,"children":187},{"id":186},"fast",[188],{"type":41,"tag":56,"props":189,"children":191},{"className":190},[],[192],{"type":47,"value":186},{"type":41,"tag":112,"props":194,"children":195},{},[196,205,214,223,232,241],{"type":41,"tag":116,"props":197,"children":198},{},[199,203],{"type":41,"tag":120,"props":200,"children":201},{},[202],{"type":47,"value":124},{"type":47,"value":204}," Quick utility tasks — parsing, classification, file ops, bulk work",{"type":41,"tag":116,"props":206,"children":207},{},[208,212],{"type":41,"tag":120,"props":209,"children":210},{},[211],{"type":47,"value":134},{"type":47,"value":213}," Cheap (Haiku, GPT luna, Gemini Flash)",{"type":41,"tag":116,"props":215,"children":216},{},[217,221],{"type":41,"tag":120,"props":218,"children":219},{},[220],{"type":47,"value":144},{"type":47,"value":222}," Well-defined, low-ambiguity, high-volume, or latency-sensitive tasks — file operations, git commands, notification triage, data extraction",{"type":41,"tag":116,"props":224,"children":225},{},[226,230],{"type":41,"tag":120,"props":227,"children":228},{},[229],{"type":47,"value":154},{"type":47,"value":231}," If the agent needs judgment calls, synthesis, or quality prose",{"type":41,"tag":116,"props":233,"children":234},{},[235,239],{"type":41,"tag":120,"props":236,"children":237},{},[238],{"type":47,"value":164},{"type":47,"value":240}," file-ops, git-ops, shell-exec, health-checker, triage-manager",{"type":41,"tag":116,"props":242,"children":243},{},[244,248,249,254],{"type":41,"tag":120,"props":245,"children":246},{},[247],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":250,"children":252},{"className":251},[],[253],{"type":47,"value":186},{"type":47,"value":255}," (terminal — fast agents rarely need fallback)",{"type":41,"tag":257,"props":258,"children":259},"hr",{},[],{"type":41,"tag":90,"props":261,"children":263},{"id":262},"coding-domain-roles",[264],{"type":47,"value":265},"Coding Domain Roles",{"type":41,"tag":50,"props":267,"children":268},{},[269],{"type":47,"value":270},"For agents whose primary output is code.",{"type":41,"tag":102,"props":272,"children":274},{"id":273},"coding",[275],{"type":41,"tag":56,"props":276,"children":278},{"className":277},[],[279],{"type":47,"value":273},{"type":41,"tag":112,"props":281,"children":282},{},[283,292,301,310,343,352],{"type":41,"tag":116,"props":284,"children":285},{},[286,290],{"type":41,"tag":120,"props":287,"children":288},{},[289],{"type":47,"value":124},{"type":47,"value":291}," Code generation, implementation, debugging",{"type":41,"tag":116,"props":293,"children":294},{},[295,299],{"type":41,"tag":120,"props":296,"children":297},{},[298],{"type":47,"value":134},{"type":47,"value":300}," Mid, code-specialized (Sonnet, gpt-5.5, Gemini Pro)",{"type":41,"tag":116,"props":302,"children":303},{},[304,308],{"type":41,"tag":120,"props":305,"children":306},{},[307],{"type":47,"value":144},{"type":47,"value":309}," The agent writes, modifies, or debugs code as its primary activity — bug fixing, feature implementation, test writing, infrastructure-as-code",{"type":41,"tag":116,"props":311,"children":312},{},[313,317,319,325,327,333,335,341],{"type":41,"tag":120,"props":314,"children":315},{},[316],{"type":47,"value":154},{"type":47,"value":318}," If the agent primarily reviews code (use ",{"type":41,"tag":56,"props":320,"children":322},{"className":321},[],[323],{"type":47,"value":324},"critique",{"type":47,"value":326},"), designs UI layouts (use ",{"type":41,"tag":56,"props":328,"children":330},{"className":329},[],[331],{"type":47,"value":332},"ui-coding",{"type":47,"value":334},"), or audits for vulnerabilities (use ",{"type":41,"tag":56,"props":336,"children":338},{"className":337},[],[339],{"type":47,"value":340},"security-audit",{"type":47,"value":342},")",{"type":41,"tag":116,"props":344,"children":345},{},[346,350],{"type":41,"tag":120,"props":347,"children":348},{},[349],{"type":47,"value":164},{"type":47,"value":351}," bug-hunter, modular-builder, test-coverage, python-dev, rust-dev",{"type":41,"tag":116,"props":353,"children":354},{},[355,359,360],{"type":41,"tag":120,"props":356,"children":357},{},[358],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":361,"children":363},{"className":362},[],[364],{"type":47,"value":365},"[coding, general]",{"type":41,"tag":102,"props":367,"children":368},{"id":332},[369],{"type":41,"tag":56,"props":370,"children":372},{"className":371},[],[373],{"type":47,"value":332},{"type":41,"tag":112,"props":375,"children":376},{},[377,386,402,411,434,443],{"type":41,"tag":116,"props":378,"children":379},{},[380,384],{"type":41,"tag":120,"props":381,"children":382},{},[383],{"type":47,"value":124},{"type":47,"value":385}," Frontend\u002FUI code — components, layouts, styling, spatial reasoning",{"type":41,"tag":116,"props":387,"children":388},{},[389,393,395,400],{"type":41,"tag":120,"props":390,"children":391},{},[392],{"type":47,"value":134},{"type":47,"value":394}," Mid, code-specialized (same as ",{"type":41,"tag":56,"props":396,"children":398},{"className":397},[],[399],{"type":47,"value":273},{"type":47,"value":401}," today — will diverge as visually-tuned models emerge)",{"type":41,"tag":116,"props":403,"children":404},{},[405,409],{"type":41,"tag":120,"props":406,"children":407},{},[408],{"type":47,"value":144},{"type":47,"value":410}," The agent builds user-facing interfaces — components, layouts, responsive design, CSS, accessibility",{"type":41,"tag":116,"props":412,"children":413},{},[414,418,420,425,427,433],{"type":41,"tag":120,"props":415,"children":416},{},[417],{"type":47,"value":154},{"type":47,"value":419}," If the agent writes backend code that serves a frontend (use ",{"type":41,"tag":56,"props":421,"children":423},{"className":422},[],[424],{"type":47,"value":273},{"type":47,"value":426},"). If it evaluates design aesthetics without writing code (use ",{"type":41,"tag":56,"props":428,"children":430},{"className":429},[],[431],{"type":47,"value":432},"creative",{"type":47,"value":342},{"type":41,"tag":116,"props":435,"children":436},{},[437,441],{"type":41,"tag":120,"props":438,"children":439},{},[440],{"type":47,"value":164},{"type":47,"value":442}," component-designer, layout-architect, responsive-strategist",{"type":41,"tag":116,"props":444,"children":445},{},[446,450,451],{"type":41,"tag":120,"props":447,"children":448},{},[449],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":452,"children":454},{"className":453},[],[455],{"type":47,"value":456},"[ui-coding, coding, general]",{"type":41,"tag":102,"props":458,"children":459},{"id":340},[460],{"type":41,"tag":56,"props":461,"children":463},{"className":462},[],[464],{"type":47,"value":340},{"type":41,"tag":112,"props":466,"children":467},{},[468,477,486,495,517,526],{"type":41,"tag":116,"props":469,"children":470},{},[471,475],{"type":41,"tag":120,"props":472,"children":473},{},[474],{"type":47,"value":124},{"type":47,"value":476}," Vulnerability assessment, attack surface analysis, code auditing",{"type":41,"tag":116,"props":478,"children":479},{},[480,484],{"type":41,"tag":120,"props":481,"children":482},{},[483],{"type":47,"value":134},{"type":47,"value":485}," Mid, code-specialized + high reasoning",{"type":41,"tag":116,"props":487,"children":488},{},[489,493],{"type":41,"tag":120,"props":490,"children":491},{},[492],{"type":47,"value":144},{"type":47,"value":494}," The agent examines code or systems for security vulnerabilities, reviews authentication flows, assesses attack surfaces",{"type":41,"tag":116,"props":496,"children":497},{},[498,502,504,509,511,516],{"type":41,"tag":120,"props":499,"children":500},{},[501],{"type":47,"value":154},{"type":47,"value":503}," If the agent is doing general code review (use ",{"type":41,"tag":56,"props":505,"children":507},{"className":506},[],[508],{"type":47,"value":324},{"type":47,"value":510},"). If it's implementing security features (use ",{"type":41,"tag":56,"props":512,"children":514},{"className":513},[],[515],{"type":47,"value":273},{"type":47,"value":342},{"type":41,"tag":116,"props":518,"children":519},{},[520,524],{"type":41,"tag":120,"props":521,"children":522},{},[523],{"type":47,"value":164},{"type":47,"value":525}," security-guardian",{"type":41,"tag":116,"props":527,"children":528},{},[529,533,534],{"type":41,"tag":120,"props":530,"children":531},{},[532],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":535,"children":537},{"className":536},[],[538],{"type":47,"value":539},"[security-audit, critique, general]",{"type":41,"tag":257,"props":541,"children":542},{},[],{"type":41,"tag":90,"props":544,"children":546},{"id":545},"cognitive-mode-roles",[547],{"type":47,"value":548},"Cognitive Mode Roles",{"type":41,"tag":50,"props":550,"children":551},{},[552],{"type":47,"value":553},"For agents whose primary value is in HOW they think.",{"type":41,"tag":102,"props":555,"children":557},{"id":556},"reasoning",[558],{"type":41,"tag":56,"props":559,"children":561},{"className":560},[],[562],{"type":47,"value":556},{"type":41,"tag":112,"props":564,"children":565},{},[566,575,584,593,623,632],{"type":41,"tag":116,"props":567,"children":568},{},[569,573],{"type":41,"tag":120,"props":570,"children":571},{},[572],{"type":47,"value":124},{"type":47,"value":574}," Deep architectural reasoning, system design, complex multi-step analysis",{"type":41,"tag":116,"props":576,"children":577},{},[578,582],{"type":41,"tag":120,"props":579,"children":580},{},[581],{"type":47,"value":134},{"type":47,"value":583}," Heavy (Opus, gpt-5.5-pro, Gemini Pro) with high reasoning effort",{"type":41,"tag":116,"props":585,"children":586},{},[587,591],{"type":41,"tag":120,"props":588,"children":589},{},[590],{"type":47,"value":144},{"type":47,"value":592}," The agent designs systems, plans architectures, breaks complex problems into steps, or needs extended chain-of-thought",{"type":41,"tag":116,"props":594,"children":595},{},[596,600,602,607,609,615,617,622],{"type":41,"tag":120,"props":597,"children":598},{},[599],{"type":47,"value":154},{"type":47,"value":601}," If the agent evaluates existing work (use ",{"type":41,"tag":56,"props":603,"children":605},{"className":604},[],[606],{"type":47,"value":324},{"type":47,"value":608},"), writes prose (use ",{"type":41,"tag":56,"props":610,"children":612},{"className":611},[],[613],{"type":47,"value":614},"writing",{"type":47,"value":616},"), or generates creative concepts (use ",{"type":41,"tag":56,"props":618,"children":620},{"className":619},[],[621],{"type":47,"value":432},{"type":47,"value":342},{"type":41,"tag":116,"props":624,"children":625},{},[626,630],{"type":41,"tag":120,"props":627,"children":628},{},[629],{"type":47,"value":164},{"type":47,"value":631}," zen-architect, brainstormer, plan-writer, recipe-author",{"type":41,"tag":116,"props":633,"children":634},{},[635,639,640],{"type":41,"tag":120,"props":636,"children":637},{},[638],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":641,"children":643},{"className":642},[],[644],{"type":47,"value":645},"[reasoning, general]",{"type":41,"tag":102,"props":647,"children":648},{"id":324},[649],{"type":41,"tag":56,"props":650,"children":652},{"className":651},[],[653],{"type":47,"value":324},{"type":41,"tag":112,"props":655,"children":656},{},[657,666,675,684,713,722],{"type":41,"tag":116,"props":658,"children":659},{},[660,664],{"type":41,"tag":120,"props":661,"children":662},{},[663],{"type":47,"value":124},{"type":47,"value":665}," Analytical evaluation — finding flaws in existing work, not generating solutions",{"type":41,"tag":116,"props":667,"children":668},{},[669,673],{"type":41,"tag":120,"props":670,"children":671},{},[672],{"type":47,"value":134},{"type":47,"value":674}," Mid with extra-high reasoning effort",{"type":41,"tag":116,"props":676,"children":677},{},[678,682],{"type":41,"tag":120,"props":679,"children":680},{},[681],{"type":47,"value":144},{"type":47,"value":683}," The agent reviews, evaluates, or finds flaws in existing code, architecture, or plans",{"type":41,"tag":116,"props":685,"children":686},{},[687,691,693,698,700,705,707,712],{"type":41,"tag":120,"props":688,"children":689},{},[690],{"type":47,"value":154},{"type":47,"value":692}," If the agent builds something new (use ",{"type":41,"tag":56,"props":694,"children":696},{"className":695},[],[697],{"type":47,"value":556},{"type":47,"value":699}," or ",{"type":41,"tag":56,"props":701,"children":703},{"className":702},[],[704],{"type":47,"value":273},{"type":47,"value":706},"). If it checks for security vulnerabilities specifically (use ",{"type":41,"tag":56,"props":708,"children":710},{"className":709},[],[711],{"type":47,"value":340},{"type":47,"value":342},{"type":41,"tag":116,"props":714,"children":715},{},[716,720],{"type":41,"tag":120,"props":717,"children":718},{},[719],{"type":47,"value":164},{"type":47,"value":721}," spec-reviewer, code-quality-reviewer, friction-detector",{"type":41,"tag":116,"props":723,"children":724},{},[725,729,730,736,737],{"type":41,"tag":120,"props":726,"children":727},{},[728],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":731,"children":733},{"className":732},[],[734],{"type":47,"value":735},"[critique, reasoning, general]",{"type":47,"value":699},{"type":41,"tag":56,"props":738,"children":740},{"className":739},[],[741],{"type":47,"value":742},"[critique, general]",{"type":41,"tag":102,"props":744,"children":745},{"id":432},[746],{"type":41,"tag":56,"props":747,"children":749},{"className":748},[],[750],{"type":47,"value":432},{"type":41,"tag":112,"props":752,"children":753},{},[754,763,772,781,803,812],{"type":41,"tag":116,"props":755,"children":756},{},[757,761],{"type":41,"tag":120,"props":758,"children":759},{},[760],{"type":47,"value":124},{"type":47,"value":762}," Design direction, aesthetic judgment, high-quality creative output",{"type":41,"tag":116,"props":764,"children":765},{},[766,770],{"type":41,"tag":120,"props":767,"children":768},{},[769],{"type":47,"value":134},{"type":47,"value":771}," Heavy (Opus, gpt-5.5, Gemini Pro)",{"type":41,"tag":116,"props":773,"children":774},{},[775,779],{"type":41,"tag":120,"props":776,"children":777},{},[778],{"type":47,"value":144},{"type":47,"value":780}," The agent makes aesthetic judgments, establishes design direction, creates visual concepts",{"type":41,"tag":116,"props":782,"children":783},{},[784,788,790,795,797,802],{"type":41,"tag":120,"props":785,"children":786},{},[787],{"type":47,"value":154},{"type":47,"value":789}," If the agent writes long-form content (use ",{"type":41,"tag":56,"props":791,"children":793},{"className":792},[],[794],{"type":47,"value":614},{"type":47,"value":796},"). If it writes UI code (use ",{"type":41,"tag":56,"props":798,"children":800},{"className":799},[],[801],{"type":47,"value":332},{"type":47,"value":342},{"type":41,"tag":116,"props":804,"children":805},{},[806,810],{"type":41,"tag":120,"props":807,"children":808},{},[809],{"type":47,"value":164},{"type":47,"value":811}," art-director, style-curator, storyboard-writer, character-designer",{"type":41,"tag":116,"props":813,"children":814},{},[815,819,820,826,827],{"type":41,"tag":120,"props":816,"children":817},{},[818],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":821,"children":823},{"className":822},[],[824],{"type":47,"value":825},"[creative, general]",{"type":47,"value":699},{"type":41,"tag":56,"props":828,"children":830},{"className":829},[],[831],{"type":47,"value":832},"[creative, writing, general]",{"type":41,"tag":102,"props":834,"children":835},{"id":614},[836],{"type":41,"tag":56,"props":837,"children":839},{"className":838},[],[840],{"type":47,"value":614},{"type":41,"tag":112,"props":842,"children":843},{},[844,853,861,870,892,901],{"type":41,"tag":116,"props":845,"children":846},{},[847,851],{"type":41,"tag":120,"props":848,"children":849},{},[850],{"type":47,"value":124},{"type":47,"value":852}," Long-form content — documentation, marketing, case studies, storytelling",{"type":41,"tag":116,"props":854,"children":855},{},[856,860],{"type":41,"tag":120,"props":857,"children":858},{},[859],{"type":47,"value":134},{"type":47,"value":771},{"type":41,"tag":116,"props":862,"children":863},{},[864,868],{"type":41,"tag":120,"props":865,"children":866},{},[867],{"type":47,"value":144},{"type":47,"value":869}," The agent produces sustained written content — documentation, blog posts, case studies, release notes",{"type":41,"tag":116,"props":871,"children":872},{},[873,877,879,884,886,891],{"type":41,"tag":120,"props":874,"children":875},{},[876],{"type":47,"value":154},{"type":47,"value":878}," If the output is code with comments (use ",{"type":41,"tag":56,"props":880,"children":882},{"className":881},[],[883],{"type":47,"value":273},{"type":47,"value":885},"). If the writing is short utility text (use ",{"type":41,"tag":56,"props":887,"children":889},{"className":888},[],[890],{"type":47,"value":186},{"type":47,"value":342},{"type":41,"tag":116,"props":893,"children":894},{},[895,899],{"type":41,"tag":120,"props":896,"children":897},{},[898],{"type":47,"value":164},{"type":47,"value":900}," technical-writer, marketing-writer, release-manager",{"type":41,"tag":116,"props":902,"children":903},{},[904,908,909,915,916],{"type":41,"tag":120,"props":905,"children":906},{},[907],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":910,"children":912},{"className":911},[],[913],{"type":47,"value":914},"[writing, creative, general]",{"type":47,"value":699},{"type":41,"tag":56,"props":917,"children":919},{"className":918},[],[920],{"type":47,"value":921},"[writing, general]",{"type":41,"tag":102,"props":923,"children":925},{"id":924},"research",[926],{"type":41,"tag":56,"props":927,"children":929},{"className":928},[],[930],{"type":47,"value":924},{"type":41,"tag":112,"props":932,"children":933},{},[934,943,951,960,975,984],{"type":41,"tag":116,"props":935,"children":936},{},[937,941],{"type":41,"tag":120,"props":938,"children":939},{},[940],{"type":47,"value":124},{"type":47,"value":942}," Deep investigation, information synthesis across multiple sources",{"type":41,"tag":116,"props":944,"children":945},{},[946,950],{"type":41,"tag":120,"props":947,"children":948},{},[949],{"type":47,"value":134},{"type":47,"value":583},{"type":41,"tag":116,"props":952,"children":953},{},[954,958],{"type":41,"tag":120,"props":955,"children":956},{},[957],{"type":47,"value":144},{"type":47,"value":959}," The agent investigates and synthesizes from multiple sources. Extended context windows matter",{"type":41,"tag":116,"props":961,"children":962},{},[963,967,969,974],{"type":41,"tag":120,"props":964,"children":965},{},[966],{"type":47,"value":154},{"type":47,"value":968}," If the agent designs based on what it already knows (use ",{"type":41,"tag":56,"props":970,"children":972},{"className":971},[],[973],{"type":47,"value":556},{"type":47,"value":342},{"type":41,"tag":116,"props":976,"children":977},{},[978,982],{"type":41,"tag":120,"props":979,"children":980},{},[981],{"type":47,"value":164},{"type":47,"value":983}," browser-researcher, story-researcher",{"type":41,"tag":116,"props":985,"children":986},{},[987,991,992],{"type":41,"tag":120,"props":988,"children":989},{},[990],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":993,"children":995},{"className":994},[],[996],{"type":47,"value":997},"[research, general]",{"type":41,"tag":257,"props":999,"children":1000},{},[],{"type":41,"tag":90,"props":1002,"children":1004},{"id":1003},"capability-roles",[1005],{"type":47,"value":1006},"Capability Roles",{"type":41,"tag":50,"props":1008,"children":1009},{},[1010],{"type":47,"value":1011},"For agents that need specific model capabilities beyond text.",{"type":41,"tag":102,"props":1013,"children":1015},{"id":1014},"vision",[1016],{"type":41,"tag":56,"props":1017,"children":1019},{"className":1018},[],[1020],{"type":47,"value":1014},{"type":41,"tag":112,"props":1022,"children":1023},{},[1024,1033,1042,1051,1074,1083],{"type":41,"tag":116,"props":1025,"children":1026},{},[1027,1031],{"type":41,"tag":120,"props":1028,"children":1029},{},[1030],{"type":47,"value":124},{"type":47,"value":1032}," Understanding visual input — screenshots, diagrams, UI mockups",{"type":41,"tag":116,"props":1034,"children":1035},{},[1036,1040],{"type":41,"tag":120,"props":1037,"children":1038},{},[1039],{"type":47,"value":134},{"type":47,"value":1041}," Mid, multimodal (Gemini Flash, Sonnet, gpt-5.5)",{"type":41,"tag":116,"props":1043,"children":1044},{},[1045,1049],{"type":41,"tag":120,"props":1046,"children":1047},{},[1048],{"type":47,"value":144},{"type":47,"value":1050}," The agent analyzes screenshots, reads diagrams, interprets UI mockups",{"type":41,"tag":116,"props":1052,"children":1053},{},[1054,1058,1060,1066,1068,1073],{"type":41,"tag":120,"props":1055,"children":1056},{},[1057],{"type":47,"value":154},{"type":47,"value":1059}," If the agent generates images (use ",{"type":41,"tag":56,"props":1061,"children":1063},{"className":1062},[],[1064],{"type":47,"value":1065},"image-gen",{"type":47,"value":1067},"). If it writes UI code without seeing existing UI (use ",{"type":41,"tag":56,"props":1069,"children":1071},{"className":1070},[],[1072],{"type":47,"value":332},{"type":47,"value":342},{"type":41,"tag":116,"props":1075,"children":1076},{},[1077,1081],{"type":41,"tag":120,"props":1078,"children":1079},{},[1080],{"type":47,"value":164},{"type":47,"value":1082}," browser-operator, visual-documenter",{"type":41,"tag":116,"props":1084,"children":1085},{},[1086,1090,1091],{"type":41,"tag":120,"props":1087,"children":1088},{},[1089],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":1092,"children":1094},{"className":1093},[],[1095],{"type":47,"value":1096},"[vision, general]",{"type":41,"tag":102,"props":1098,"children":1099},{"id":1065},[1100],{"type":41,"tag":56,"props":1101,"children":1103},{"className":1102},[],[1104],{"type":47,"value":1065},{"type":41,"tag":112,"props":1106,"children":1107},{},[1108,1117,1126,1135,1150,1160,1169],{"type":41,"tag":116,"props":1109,"children":1110},{},[1111,1115],{"type":41,"tag":120,"props":1112,"children":1113},{},[1114],{"type":47,"value":124},{"type":47,"value":1116}," Image generation, visual mockup creation, visual ideation",{"type":41,"tag":116,"props":1118,"children":1119},{},[1120,1124],{"type":41,"tag":120,"props":1121,"children":1122},{},[1123],{"type":47,"value":134},{"type":47,"value":1125}," Specialized (gemini-3-pro-image-preview) — sparse provider coverage",{"type":41,"tag":116,"props":1127,"children":1128},{},[1129,1133],{"type":41,"tag":120,"props":1130,"children":1131},{},[1132],{"type":47,"value":144},{"type":47,"value":1134}," The agent creates images — generating mockups, producing comic panels, visual prototypes",{"type":41,"tag":116,"props":1136,"children":1137},{},[1138,1142,1144,1149],{"type":41,"tag":120,"props":1139,"children":1140},{},[1141],{"type":47,"value":154},{"type":47,"value":1143}," If the agent analyzes existing images (use ",{"type":41,"tag":56,"props":1145,"children":1147},{"className":1146},[],[1148],{"type":47,"value":1014},{"type":47,"value":342},{"type":41,"tag":116,"props":1151,"children":1152},{},[1153,1158],{"type":41,"tag":120,"props":1154,"children":1155},{},[1156],{"type":47,"value":1157},"IMPORTANT:",{"type":47,"value":1159}," Always include a non-image-gen fallback. Google-only in most matrices today.",{"type":41,"tag":116,"props":1161,"children":1162},{},[1163,1167],{"type":41,"tag":120,"props":1164,"children":1165},{},[1166],{"type":47,"value":164},{"type":47,"value":1168}," panel-artist, cover-artist",{"type":41,"tag":116,"props":1170,"children":1171},{},[1172,1176,1177,1183,1185,1190],{"type":41,"tag":120,"props":1173,"children":1174},{},[1175],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":1178,"children":1180},{"className":1179},[],[1181],{"type":47,"value":1182},"[image-gen, creative, general]",{"type":47,"value":1184}," (NEVER ",{"type":41,"tag":56,"props":1186,"children":1188},{"className":1187},[],[1189],{"type":47,"value":1065},{"type":47,"value":1191}," alone)",{"type":41,"tag":257,"props":1193,"children":1194},{},[],{"type":41,"tag":90,"props":1196,"children":1198},{"id":1197},"operational-role",[1199],{"type":47,"value":1200},"Operational Role",{"type":41,"tag":102,"props":1202,"children":1204},{"id":1203},"critical-ops",[1205],{"type":41,"tag":56,"props":1206,"children":1208},{"className":1207},[],[1209],{"type":47,"value":1203},{"type":41,"tag":112,"props":1211,"children":1212},{},[1213,1222,1231,1240,1262,1271],{"type":41,"tag":116,"props":1214,"children":1215},{},[1216,1220],{"type":41,"tag":120,"props":1217,"children":1218},{},[1219],{"type":47,"value":124},{"type":47,"value":1221}," High-reliability operational tasks — infrastructure, orchestration, coordination where mistakes are costly",{"type":41,"tag":116,"props":1223,"children":1224},{},[1225,1229],{"type":41,"tag":120,"props":1226,"children":1227},{},[1228],{"type":47,"value":134},{"type":47,"value":1230}," Heavy (Opus, gpt-5.5-pro, Gemini Pro)",{"type":41,"tag":116,"props":1232,"children":1233},{},[1234,1238],{"type":41,"tag":120,"props":1235,"children":1236},{},[1237],{"type":47,"value":144},{"type":47,"value":1239}," The agent orchestrates infrastructure, manages deployments, verifies shadow environments, or performs operational tasks where failures cascade",{"type":41,"tag":116,"props":1241,"children":1242},{},[1243,1247,1249,1254,1256,1261],{"type":41,"tag":120,"props":1244,"children":1245},{},[1246],{"type":47,"value":154},{"type":47,"value":1248}," If the agent writes IaC as its primary activity (use ",{"type":41,"tag":56,"props":1250,"children":1252},{"className":1251},[],[1253],{"type":47,"value":273},{"type":47,"value":1255},"). If it does quick operational checks (use ",{"type":41,"tag":56,"props":1257,"children":1259},{"className":1258},[],[1260],{"type":47,"value":186},{"type":47,"value":342},{"type":41,"tag":116,"props":1263,"children":1264},{},[1265,1269],{"type":41,"tag":120,"props":1266,"children":1267},{},[1268],{"type":47,"value":164},{"type":47,"value":1270}," setup-digital-twin, dtu-profile-builder, container-operator",{"type":41,"tag":116,"props":1272,"children":1273},{},[1274,1278,1279],{"type":41,"tag":120,"props":1275,"children":1276},{},[1277],{"type":47,"value":174},{"type":47,"value":176},{"type":41,"tag":56,"props":1280,"children":1282},{"className":1281},[],[1283],{"type":47,"value":1284},"[critical-ops, coding, general]",{"type":41,"tag":257,"props":1286,"children":1287},{},[],{"type":41,"tag":65,"props":1289,"children":1291},{"id":1290},"model-tier-grid",[1292],{"type":47,"value":1293},"Model Tier Grid",{"type":41,"tag":50,"props":1295,"children":1296},{},[1297],{"type":47,"value":1298},"Every role maps to a unique (model-tier x reasoning-config) cell:",{"type":41,"tag":72,"props":1300,"children":1303},{"className":1301,"code":1302,"language":47},[75],"                default reasoning        high reasoning         extra-high reasoning\n             ┌──────────────────────┬──────────────────────┬──────────────────────┐\n  Heavy      │  creative            │  reasoning           │                      │\n  (Opus)     │  writing             │  research            │                      │\n             │  critical-ops        │                      │                      │\n             ├──────────────────────┼──────────────────────┼──────────────────────┤\n  Mid        │  general   coding    │  security-audit      │  critique            │\n  (Sonnet)   │  ui-coding           │                      │                      │\n             ├──────────────────────┤                      │                      │\n  Flash      │  fast     vision     │                      │                      │\n  (Haiku)    │                      │                      │                      │\n             ├──────────────────────┤                      │                      │\n  Specialized│  image-gen           │                      │                      │\n             └──────────────────────┴──────────────────────┴──────────────────────┘\n",[1304],{"type":41,"tag":56,"props":1305,"children":1306},{"__ignoreMap":80},[1307],{"type":47,"value":1302},{"type":41,"tag":65,"props":1309,"children":1311},{"id":1310},"fallback-chain-best-practices",[1312],{"type":47,"value":1313},"Fallback Chain Best Practices",{"type":41,"tag":1315,"props":1316,"children":1317},"ol",{},[1318,1340,1362,1378,1388],{"type":41,"tag":116,"props":1319,"children":1320},{},[1321,1331,1333,1338],{"type":41,"tag":120,"props":1322,"children":1323},{},[1324,1326],{"type":47,"value":1325},"Every chain should end with ",{"type":41,"tag":56,"props":1327,"children":1329},{"className":1328},[],[1330],{"type":47,"value":104},{"type":47,"value":1332}," (or ",{"type":41,"tag":56,"props":1334,"children":1336},{"className":1335},[],[1337],{"type":47,"value":186},{"type":47,"value":1339}," for utility agents)",{"type":41,"tag":116,"props":1341,"children":1342},{},[1343,1348,1349,1354,1356],{"type":41,"tag":120,"props":1344,"children":1345},{},[1346],{"type":47,"value":1347},"Go specific → general:",{"type":47,"value":176},{"type":41,"tag":56,"props":1350,"children":1352},{"className":1351},[],[1353],{"type":47,"value":456},{"type":47,"value":1355}," not ",{"type":41,"tag":56,"props":1357,"children":1359},{"className":1358},[],[1360],{"type":47,"value":1361},"[general, ui-coding]",{"type":41,"tag":116,"props":1363,"children":1364},{},[1365,1370,1371,1376],{"type":41,"tag":120,"props":1366,"children":1367},{},[1368],{"type":47,"value":1369},"Sparse roles need fallbacks:",{"type":47,"value":176},{"type":41,"tag":56,"props":1372,"children":1374},{"className":1373},[],[1375],{"type":47,"value":1182},{"type":47,"value":1377}," — image-gen has limited provider coverage",{"type":41,"tag":116,"props":1379,"children":1380},{},[1381,1386],{"type":41,"tag":120,"props":1382,"children":1383},{},[1384],{"type":47,"value":1385},"Don't over-chain:",{"type":47,"value":1387}," 2-3 roles is typical. More than 4 suggests confusion about the agent's purpose",{"type":41,"tag":116,"props":1389,"children":1390},{},[1391,1396,1398,1403,1405,1410,1411],{"type":41,"tag":120,"props":1392,"children":1393},{},[1394],{"type":47,"value":1395},"Single role is fine",{"type":47,"value":1397}," when the agent's need is unambiguous: ",{"type":41,"tag":56,"props":1399,"children":1401},{"className":1400},[],[1402],{"type":47,"value":186},{"type":47,"value":1404},", ",{"type":41,"tag":56,"props":1406,"children":1408},{"className":1407},[],[1409],{"type":47,"value":104},{"type":47,"value":1404},{"type":41,"tag":56,"props":1412,"children":1414},{"className":1413},[],[1415],{"type":47,"value":273},{"items":1417,"total":1608},[1418,1440,1457,1478,1493,1510,1521,1534,1549,1564,1583,1596],{"slug":1419,"name":1419,"fn":1420,"description":1421,"org":1422,"tags":1423,"stars":1437,"repoUrl":1438,"updatedAt":1439},"rushstack-best-practices","manage Rush monorepos with best practices","Provides best practices and guidance for working with Rush monorepos. Use when the user is working in a Rush-based repository, asks about Rush commands (install, update, build, rebuild), needs help with project selection, dependency management, build caching, subspace configuration, or troubleshooting Rush-specific issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1424,1427,1430,1431,1434],{"name":1425,"slug":1426,"type":15},"Engineering","engineering",{"name":1428,"slug":1429,"type":15},"Local Development","local-development",{"name":9,"slug":8,"type":15},{"name":1432,"slug":1433,"type":15},"Project Management","project-management",{"name":1435,"slug":1436,"type":15},"Rush","rush",6484,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Frushstack","2026-04-06T18:34:44.965032",{"slug":1441,"name":1441,"fn":1442,"description":1443,"org":1444,"tags":1445,"stars":1454,"repoUrl":1455,"updatedAt":1456},"azure-ai-agents-persistent-dotnet","build AI agents with Azure .NET SDK","Azure AI Agents Persistent SDK for .NET. Low-level SDK for creating and managing AI agents with threads, messages, runs, and tools. Use for agent CRUD, conversation threads, streaming responses, function calling, file search, and code interpreter. Triggers: \"PersistentAgentsClient\", \"persistent agents\", \"agent threads\", \"agent runs\", \"streaming agents\", \"function calling agents .NET\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1446,1449,1450,1453],{"name":1447,"slug":1448,"type":15},".NET","net",{"name":21,"slug":22,"type":15},{"name":1451,"slug":1452,"type":15},"Azure","azure",{"name":13,"slug":14,"type":15},2804,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fskills","2026-07-03T16:32:10.297433",{"slug":1458,"name":1458,"fn":1459,"description":1460,"org":1461,"tags":1462,"stars":1454,"repoUrl":1455,"updatedAt":1477},"azure-ai-anomalydetector-java","build anomaly detection applications with Java","Build anomaly detection applications with Azure AI Anomaly Detector SDK for Java. Use when implementing univariate\u002Fmultivariate anomaly detection, time-series analysis, or AI-powered monitoring.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1463,1466,1467,1470,1473,1474],{"name":1464,"slug":1465,"type":15},"Analytics","analytics",{"name":1451,"slug":1452,"type":15},{"name":1468,"slug":1469,"type":15},"Data Analysis","data-analysis",{"name":1471,"slug":1472,"type":15},"Java","java",{"name":9,"slug":8,"type":15},{"name":1475,"slug":1476,"type":15},"Monitoring","monitoring","2026-05-13T06:14:16.261754",{"slug":1479,"name":1479,"fn":1480,"description":1481,"org":1482,"tags":1483,"stars":1454,"repoUrl":1455,"updatedAt":1492},"azure-ai-contentsafety-java","build content moderation applications with Azure AI","Build content moderation applications with Azure AI Content Safety SDK for Java. Use when implementing text\u002Fimage analysis, blocklist management, or harm detection for hate, violence, sexual content, and self-harm.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1484,1487,1488,1489],{"name":1485,"slug":1486,"type":15},"AI Infrastructure","ai-infrastructure",{"name":1451,"slug":1452,"type":15},{"name":1471,"slug":1472,"type":15},{"name":1490,"slug":1491,"type":15},"Security","security","2026-07-07T06:53:31.293235",{"slug":1494,"name":1494,"fn":1495,"description":1496,"org":1497,"tags":1498,"stars":1454,"repoUrl":1455,"updatedAt":1509},"azure-ai-contentsafety-py","detect harmful content with Azure AI Content Safety","Azure AI Content Safety SDK for Python. Use for detecting harmful content in text and images with multi-severity classification.\nTriggers: \"azure-ai-contentsafety\", \"ContentSafetyClient\", \"content moderation\", \"harmful content\", \"text analysis\", \"image analysis\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1499,1500,1503,1504,1505,1508],{"name":1451,"slug":1452,"type":15},{"name":1501,"slug":1502,"type":15},"Compliance","compliance",{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},{"name":1506,"slug":1507,"type":15},"Python","python",{"name":1490,"slug":1491,"type":15},"2026-07-18T05:14:23.017504",{"slug":1511,"name":1511,"fn":1512,"description":1513,"org":1514,"tags":1515,"stars":1454,"repoUrl":1455,"updatedAt":1520},"azure-ai-language-conversations-py","implement conversational language understanding with Python","Implement Conversational Language Understanding (CLU) using the azure-ai-language-conversations Python SDK. Use when working with ConversationAnalysisClient to analyze conversation intent and entities, building NLP features, or integrating language understanding into applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1516,1517,1518,1519],{"name":1464,"slug":1465,"type":15},{"name":1451,"slug":1452,"type":15},{"name":13,"slug":14,"type":15},{"name":1506,"slug":1507,"type":15},"2026-07-31T05:54:29.068751",{"slug":1522,"name":1522,"fn":1523,"description":1524,"org":1525,"tags":1526,"stars":1454,"repoUrl":1455,"updatedAt":1533},"azure-ai-translation-text-py","translate text using Azure AI services","Azure AI Text Translation SDK for real-time text translation, transliteration, language detection, and dictionary lookup. Use for translating text content in applications.\nTriggers: \"text translation\", \"translator\", \"translate text\", \"transliterate\", \"TextTranslationClient\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1527,1530,1531,1532],{"name":1528,"slug":1529,"type":15},"API Development","api-development",{"name":1451,"slug":1452,"type":15},{"name":9,"slug":8,"type":15},{"name":1506,"slug":1507,"type":15},"2026-07-18T05:14:16.988376",{"slug":1535,"name":1535,"fn":1536,"description":1537,"org":1538,"tags":1539,"stars":1454,"repoUrl":1455,"updatedAt":1548},"azure-ai-vision-imageanalysis-py","analyze images with Azure AI Vision","Azure AI Vision Image Analysis SDK for captions, tags, objects, OCR, people detection, and smart cropping. Use for computer vision and image understanding tasks.\nTriggers: \"image analysis\", \"computer vision\", \"OCR\", \"object detection\", \"ImageAnalysisClient\", \"image caption\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1540,1541,1544,1547],{"name":1451,"slug":1452,"type":15},{"name":1542,"slug":1543,"type":15},"Computer Vision","computer-vision",{"name":1545,"slug":1546,"type":15},"Images","images",{"name":1506,"slug":1507,"type":15},"2026-07-18T05:14:18.007737",{"slug":1550,"name":1550,"fn":1551,"description":1552,"org":1553,"tags":1554,"stars":1454,"repoUrl":1455,"updatedAt":1563},"azure-appconfiguration-java","manage configuration with Azure App Configuration","Azure App Configuration SDK for Java. Centralized application configuration management with key-value settings, feature flags, and snapshots.\nTriggers: \"ConfigurationClient java\", \"app configuration java\", \"feature flag java\", \"configuration setting java\", \"azure config java\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1555,1556,1559,1562],{"name":1451,"slug":1452,"type":15},{"name":1557,"slug":1558,"type":15},"Configuration","configuration",{"name":1560,"slug":1561,"type":15},"Feature Flags","feature-flags",{"name":1471,"slug":1472,"type":15},"2026-07-03T16:32:01.278468",{"slug":1565,"name":1565,"fn":1566,"description":1567,"org":1568,"tags":1569,"stars":1454,"repoUrl":1455,"updatedAt":1582},"azure-cosmos-rust","build applications with Azure Cosmos DB","Azure Cosmos DB library for Rust (NoSQL API). Document CRUD, containers, and globally distributed data.\nTriggers: \"cosmos db rust\", \"CosmosClient rust\", \"document crud rust\", \"NoSQL rust\", \"partition key rust\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1570,1573,1576,1579],{"name":1571,"slug":1572,"type":15},"Cosmos DB","cosmos-db",{"name":1574,"slug":1575,"type":15},"Database","database",{"name":1577,"slug":1578,"type":15},"NoSQL","nosql",{"name":1580,"slug":1581,"type":15},"Rust","rust","2026-07-31T05:54:27.021432",{"slug":1584,"name":1584,"fn":1566,"description":1585,"org":1586,"tags":1587,"stars":1454,"repoUrl":1455,"updatedAt":1595},"azure-cosmos-ts","Azure Cosmos DB JavaScript\u002FTypeScript SDK (@azure\u002Fcosmos) for data plane operations. Use for CRUD operations on documents, queries, bulk operations, and container management. Triggers: \"Cosmos DB\", \"@azure\u002Fcosmos\", \"CosmosClient\", \"document CRUD\", \"NoSQL queries\", \"bulk operations\", \"partition key\", \"container.items\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1588,1589,1590,1591,1592],{"name":1571,"slug":1572,"type":15},{"name":1574,"slug":1575,"type":15},{"name":9,"slug":8,"type":15},{"name":1577,"slug":1578,"type":15},{"name":1593,"slug":1594,"type":15},"TypeScript","typescript","2026-07-03T16:31:19.368382",{"slug":1597,"name":1597,"fn":1598,"description":1599,"org":1600,"tags":1601,"stars":1454,"repoUrl":1455,"updatedAt":1607},"azure-data-tables-java","build table storage applications with Java","Build table storage applications with Azure Tables SDK for Java. Use when working with Azure Table Storage or Cosmos DB Table API for NoSQL key-value data, schemaless storage, or structured data at scale.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1602,1603,1604,1605,1606],{"name":1451,"slug":1452,"type":15},{"name":1571,"slug":1572,"type":15},{"name":1574,"slug":1575,"type":15},{"name":1471,"slug":1472,"type":15},{"name":1577,"slug":1578,"type":15},"2026-05-13T06:14:17.582229",267,{"items":1610,"total":1618},[1611],{"slug":4,"name":4,"fn":5,"description":6,"org":1612,"tags":1613,"stars":23,"repoUrl":24,"updatedAt":25},{"slug":8,"name":9,"logoUrl":10,"githubOrg":8},[1614,1615,1616,1617],{"name":21,"slug":22,"type":15},{"name":17,"slug":18,"type":15},{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},1]