Core Capability Matrix
Development Tasks
| Task Type |
Primary Agent |
Fallback |
Model |
Notes |
| Debug complex issue |
debug-like-expert skill |
general-purpose |
sonnet |
Hypothesis-driven |
| Quick bug fix |
general-purpose |
— |
sonnet |
Simple fixes |
| Code review |
code-reviewer |
feature-dev:code-reviewer |
haiku |
Pattern matching |
| Architecture design |
Plan agent |
code-architect |
opus |
Complex reasoning |
| Feature implementation |
general-purpose |
feature-dev agents |
sonnet |
Code generation |
| Refactoring |
code-simplifier |
general-purpose |
sonnet |
Preserve behavior |
| Security audit |
security skill |
general-purpose |
sonnet |
OWASP patterns |
| Write tests |
testing skill |
general-purpose |
sonnet |
TDD workflow |
| API design |
api-design skill |
general-purpose |
sonnet |
REST/GraphQL |
| API testing |
api-testing skill |
general-purpose |
haiku |
Postman/Bruno |
Exploration & Research
| Task Type |
Primary Agent |
Fallback |
Model |
Notes |
| Find files by pattern |
Glob tool (direct) |
Explore agent |
— |
Fastest |
| Search code content |
Grep tool (direct) |
Explore agent |
— |
Fastest |
| Understand codebase area |
Explore agent |
general-purpose |
haiku |
Thorough search |
| Deep code analysis |
code-explorer |
Explore agent |
sonnet |
Traces execution |
| Research framework/tool |
research skill |
WebSearch |
sonnet |
Structured eval |
| Market research |
research skill |
WebSearch |
sonnet |
Company profiles |
Infrastructure & Deployment
| Task Type |
Primary Agent |
Fallback |
Model |
Notes |
| Docker setup |
docker-compose skill |
general-purpose |
sonnet |
Local dev |
| GPU deployment |
runpod-deployment skill |
general-purpose |
sonnet |
RunPod |
| Database migration |
supabase-sql skill |
general-purpose |
sonnet |
Supabase |
| Stripe integration |
stripe-stack skill |
general-purpose |
sonnet |
Payments |
| Voice AI pipeline |
voice-ai skill |
general-purpose |
sonnet |
Deepgram+Cartesia |
| LangGraph agents |
langgraph-agents skill |
general-purpose |
sonnet |
Multi-agent |
| Fast inference |
groq-inference skill |
general-purpose |
sonnet |
GROQ API |
| Chinese LLMs |
openrouter skill |
general-purpose |
sonnet |
DeepSeek/Qwen |
| LLM fine-tuning |
unsloth-training skill |
general-purpose |
sonnet |
GRPO/SFT |
Business & GTM
| Task Type |
Primary Agent |
Fallback |
Model |
Notes |
| GTM strategy |
gtm-pricing skill |
general-purpose |
sonnet |
ICP, positioning |
| Sales outreach |
sales-revenue skill |
general-purpose |
sonnet |
Cold email |
| CRM integration |
crm-integration skill |
general-purpose |
sonnet |
Close/HubSpot |
| Content marketing |
content-marketing skill |
general-purpose |
sonnet |
B2B content |
| Data analysis |
data-analysis skill |
general-purpose |
sonnet |
Dashboards |
| Trading signals |
trading-signals skill |
general-purpose |
sonnet |
Technical analysis |
| Business model |
business-model-canvas skill |
general-purpose |
sonnet |
9-block canvas |
| Market differentiation |
blue-ocean-strategy skill |
general-purpose |
sonnet |
ERRC framework |
Workflow & Coordination
| Task Type |
Primary Agent |
Fallback |
Model |
Notes |
| Session start/end |
workflow-orchestrator skill |
— |
sonnet |
Daily lifecycle |
| Parallel worktree dev |
agent-teams skill |
worktree-manager skill |
sonnet |
Full isolation |
| In-session parallel |
subagent-teams skill |
sequential execution |
haiku-sonnet |
Task tool |
| Progress tracking |
TaskCreate/TaskUpdate |
TodoWrite |
— |
Native UI spinners |
| Team coordination |
TeamCreate + SendMessage |
subagent-teams skill |
haiku-sonnet |
Shared task list |
| Git workflow |
git-workflow skill |
general-purpose |
haiku |
Commits, PRs |
| Skill authoring |
extension-authoring skill |
general-purpose |
sonnet |
Skills, hooks |
| Project planning |
planning-prompts skill |
Plan agent |
sonnet |
Meta-prompts |
| Cost tracking |
cost-metering skill |
manual tracking |
haiku |
Budget status |
Decision Flowchart
START: What do I need to do?
│
├── Search/Find something?
│ ├── Know the file pattern? → Glob (direct)
│ ├── Know the code pattern? → Grep (direct)
│ └── Need to understand area? → Explore agent (haiku)
│
├── Write/Change code?
│ ├── Simple fix? → general-purpose (sonnet)
│ ├── New feature? → Check if a skill matches → skill or general-purpose (sonnet)
│ ├── Architecture? → Plan agent (opus)
│ └── Parallel build? → subagent-teams or agent-teams
│
├── Review/Validate?
│ ├── Code review? → code-reviewer (haiku)
│ ├── Security? → security skill (sonnet)
│ └── Tests? → testing skill (sonnet)
│
├── Research/Learn?
│ ├── Codebase? → Explore agent (haiku)
│ ├── Framework? → research skill (sonnet)
│ └── Market? → research skill (sonnet)
│
└── Business/GTM?
└── Check Business section of matrix above
Model Tier Guide
| Tier |
Model |
Cost/1M In |
Best For |
| Fast |
Haiku 4.5 |
$1.00 |
Search, classify, review, simple tasks |
| Standard |
Sonnet 4.6 |
$3.00 |
Code generation, reasoning, most tasks |
| Premium |
Opus 4.6 |
$5.00 |
Architecture, complex decisions, planning |
| Free |
TaskCreate |
$0 |
Progress tracking (local UI, not an API call) |
Default to Sonnet unless:
- Task is search/classify → use Haiku (5x cheaper)
- Task requires deep reasoning → use Opus (5x more capable)
- Task is progress tracking → use TaskCreate with
activeForm for live spinners (zero cost)
Cross-references: subagent-teams for in-session parallel agents, agent-teams for worktree-isolated parallel agents.
Deep dive: See reference/matrix-table.md, reference/selection-flowchart.md
Emit Outcome Sidecar
As the final step, write to ~/.claude/skill-analytics/last-outcome-agent-capability-matrix.json:
{"ts":"[UTC ISO8601]","skill":"agent-capability-matrix","version":"1.0.0","variant":"default",
"status":"[success|partial|error]","runtime_ms":[estimated ms from start],
"metrics":{"tasks_routed":[n],"agents_matched":[n],"fallbacks_used":[n]},
"error":null,"session_id":"[YYYY-MM-DD]"}
Use status "partial" if some stages failed but results were produced. Use "error" only if no output was generated.
1---2name: agent-capability-matrix3description: Map task types to the best agent, skill, model, and fallback. Route any task to the right tool. Use when: which agent, route task, agent for this, best agent, capability matrix.4---5
6<objective>
7Provide a comprehensive mapping from task types to the optimal agent, skill, model tier, and fallback strategy. Eliminates guesswork when choosing which agent to use for a given task.
8</objective>
9
10<quick_start>
11**Find the right agent:**
121. Identify your task type (debug, review, build, explore, etc.)
132. Look up the primary agent in the matrix below
143. If primary fails, use the fallback
154. Match model tier to task complexity
16</quick_start>
17
18<success_criteria>
19- Task routed to the correct primary agent on first attempt
20- Fallback agent identified if primary is unavailable or insufficient
21- Model tier matched to task complexity (Haiku for search, Sonnet for code, Opus for architecture)
22- Zero-cost tools (TaskCreate, TeamCreate) used for progress tracking instead of API calls
23- Agent selection justified by the matrix rather than ad-hoc guessing
24</success_criteria>
25
26<triggers>
27- "which agent", "route task", "agent for this", "best agent"
28- "capability matrix", "agent selection", "what tool should I use"
29- "task routing", "agent routing"
30</triggers>
31
32---
33
34## Core Capability Matrix
35
36### Development Tasks
37
38| Task Type | Primary Agent | Fallback | Model | Notes |
39|-----------|--------------|----------|-------|-------|
40| Debug complex issue | debug-like-expert skill | general-purpose | sonnet | Hypothesis-driven |
41| Quick bug fix | general-purpose | — | sonnet | Simple fixes |
42| Code review | code-reviewer | feature-dev:code-reviewer | haiku | Pattern matching |
43| Architecture design | Plan agent | code-architect | opus | Complex reasoning |
44| Feature implementation | general-purpose | feature-dev agents | sonnet | Code generation |
45| Refactoring | code-simplifier | general-purpose | sonnet | Preserve behavior |
46| Security audit | security skill | general-purpose | sonnet | OWASP patterns |
47| Write tests | testing skill | general-purpose | sonnet | TDD workflow |
48| API design | api-design skill | general-purpose | sonnet | REST/GraphQL |
49| API testing | api-testing skill | general-purpose | haiku | Postman/Bruno |
50
51### Exploration & Research
52
53| Task Type | Primary Agent | Fallback | Model | Notes |
54|-----------|--------------|----------|-------|-------|
55| Find files by pattern | Glob tool (direct) | Explore agent | — | Fastest |
56| Search code content | Grep tool (direct) | Explore agent | — | Fastest |
57| Understand codebase area | Explore agent | general-purpose | haiku | Thorough search |
58| Deep code analysis | code-explorer | Explore agent | sonnet | Traces execution |
59| Research framework/tool | research skill | WebSearch | sonnet | Structured eval |
60| Market research | research skill | WebSearch | sonnet | Company profiles |
61
62### Infrastructure & Deployment
63
64| Task Type | Primary Agent | Fallback | Model | Notes |
65|-----------|--------------|----------|-------|-------|
66| Docker setup | docker-compose skill | general-purpose | sonnet | Local dev |
67| GPU deployment | runpod-deployment skill | general-purpose | sonnet | RunPod |
68| Database migration | supabase-sql skill | general-purpose | sonnet | Supabase |
69| Stripe integration | stripe-stack skill | general-purpose | sonnet | Payments |
70| Voice AI pipeline | voice-ai skill | general-purpose | sonnet | Deepgram+Cartesia |
71| LangGraph agents | langgraph-agents skill | general-purpose | sonnet | Multi-agent |
72| Fast inference | groq-inference skill | general-purpose | sonnet | GROQ API |
73| Chinese LLMs | openrouter skill | general-purpose | sonnet | DeepSeek/Qwen |
74| LLM fine-tuning | unsloth-training skill | general-purpose | sonnet | GRPO/SFT |
75
76### Business & GTM
77
78| Task Type | Primary Agent | Fallback | Model | Notes |
79|-----------|--------------|----------|-------|-------|
80| GTM strategy | gtm-pricing skill | general-purpose | sonnet | ICP, positioning |
81| Sales outreach | sales-revenue skill | general-purpose | sonnet | Cold email |
82| CRM integration | crm-integration skill | general-purpose | sonnet | Close/HubSpot |
83| Content marketing | content-marketing skill | general-purpose | sonnet | B2B content |
84| Data analysis | data-analysis skill | general-purpose | sonnet | Dashboards |
85| Trading signals | trading-signals skill | general-purpose | sonnet | Technical analysis |
86| Business model | business-model-canvas skill | general-purpose | sonnet | 9-block canvas |
87| Market differentiation | blue-ocean-strategy skill | general-purpose | sonnet | ERRC framework |
88
89### Workflow & Coordination
90
91| Task Type | Primary Agent | Fallback | Model | Notes |
92|-----------|--------------|----------|-------|-------|
93| Session start/end | workflow-orchestrator skill | — | sonnet | Daily lifecycle |
94| Parallel worktree dev | agent-teams skill | worktree-manager skill | sonnet | Full isolation |
95| In-session parallel | subagent-teams skill | sequential execution | haiku-sonnet | Task tool |
96| Progress tracking | TaskCreate/TaskUpdate | TodoWrite | — | Native UI spinners |
97| Team coordination | TeamCreate + SendMessage | subagent-teams skill | haiku-sonnet | Shared task list |
98| Git workflow | git-workflow skill | general-purpose | haiku | Commits, PRs |
99| Skill authoring | extension-authoring skill | general-purpose | sonnet | Skills, hooks |
100| Project planning | planning-prompts skill | Plan agent | sonnet | Meta-prompts |
101| Cost tracking | cost-metering skill | manual tracking | haiku | Budget status |
102
103---
104
105## Decision Flowchart
106
107```
108START: What do I need to do?
109│
110├── Search/Find something?
111│ ├── Know the file pattern? → Glob (direct)
112│ ├── Know the code pattern? → Grep (direct)
113│ └── Need to understand area? → Explore agent (haiku)
114│
115├── Write/Change code?
116│ ├── Simple fix? → general-purpose (sonnet)
117│ ├── New feature? → Check if a skill matches → skill or general-purpose (sonnet)
118│ ├── Architecture? → Plan agent (opus)
119│ └── Parallel build? → subagent-teams or agent-teams
120│
121├── Review/Validate?
122│ ├── Code review? → code-reviewer (haiku)
123│ ├── Security? → security skill (sonnet)
124│ └── Tests? → testing skill (sonnet)
125│
126├── Research/Learn?
127│ ├── Codebase? → Explore agent (haiku)
128│ ├── Framework? → research skill (sonnet)
129│ └── Market? → research skill (sonnet)
130│
131└── Business/GTM?
132 └── Check Business section of matrix above
133```
134
135---
136
137## Model Tier Guide
138
139| Tier | Model | Cost/1M In | Best For |
140|------|-------|-----------|----------|
141| **Fast** | Haiku 4.5 | $1.00 | Search, classify, review, simple tasks |
142| **Standard** | Sonnet 4.6 | $3.00 | Code generation, reasoning, most tasks |
143| **Premium** | Opus 4.6 | $5.00 | Architecture, complex decisions, planning |
144| **Free** | TaskCreate | $0 | Progress tracking (local UI, not an API call) |
145
146**Default to Sonnet** unless:
147- Task is search/classify → use Haiku (5x cheaper)
148- Task requires deep reasoning → use Opus (5x more capable)
149- Task is progress tracking → use TaskCreate with `activeForm` for live spinners (zero cost)
150
151**Cross-references:** [subagent-teams](../subagent-teams-skill/SKILL.md) for in-session parallel agents, [agent-teams](../agent-teams-skill/SKILL.md) for worktree-isolated parallel agents.
152
153**Deep dive:** See `reference/matrix-table.md`, `reference/selection-flowchart.md`
154
155## Emit Outcome Sidecar
156
157As the final step, write to `~/.claude/skill-analytics/last-outcome-agent-capability-matrix.json`:
158```json
159{"ts":"[UTC ISO8601]","skill":"agent-capability-matrix","version":"1.0.0","variant":"default",
160 "status":"[success|partial|error]","runtime_ms":[estimated ms from start],
161 "metrics":{"tasks_routed":[n],"agents_matched":[n],"fallbacks_used":[n]},
162 "error":null,"session_id":"[YYYY-MM-DD]"}
163```
164Use status "partial" if some stages failed but results were produced. Use "error" only if no output was generated.
scientiacapital