From threadline-ai
Use this skill when the user wants to build a multi-agent system, design an agent pipeline, pick the right agentic pattern, create an orchestration workflow, or implement patterns like coordinator, swarm, or ReAct. Covers single-agent, sequential, parallel, loop, review-critique, iterative-refinement, coordinator, hierarchical, swarm, ReAct, human-in-the-loop, and custom-logic patterns. Do NOT use for tweaking a single existing agent or composing existing skills.
How this skill is triggered — by the user, by Claude, or both
Slash command
/threadline-ai:agentic-patternsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Build agentic AI systems using proven design patterns. Guides you from pattern selection through component design to working implementation using Claude Code's Agent tool.
Build agentic AI systems using proven design patterns. Guides you from pattern selection through component design to working implementation using Claude Code's Agent tool.
skill-creator:skill-creator)agent-factory)digraph patterns {
rankdir=TB;
node [shape=box];
subgraph cluster_single {
label="Single Agent";
"single-agent" [label="Single Agent\n─────────\nOne model + tools\nSimplest pattern"];
}
subgraph cluster_multi {
label="Multi-Agent Patterns";
style=dashed;
subgraph cluster_flow {
label="Flow Control";
"sequential" [label="Sequential\n─────────\nLinear pipeline\nA → B → C"];
"parallel" [label="Parallel\n─────────\nConcurrent fan-out\nA + B + C → merge"];
"loop" [label="Loop\n─────────\nRepeat until\nexit condition"];
}
subgraph cluster_quality {
label="Quality Patterns";
"review-critique" [label="Review & Critique\n─────────\nGenerator → Critic\nApprove/reject cycle"];
"iterative-refinement" [label="Iterative Refinement\n─────────\nProgressive improvement\nover multiple cycles"];
}
subgraph cluster_orchestration {
label="Orchestration";
"coordinator" [label="Coordinator\n─────────\nAI-driven routing\nto specialists"];
"hierarchical" [label="Hierarchical\n─────────\nTask decomposition\nthrough layers"];
"swarm" [label="Swarm\n─────────\nAll-to-all debate\nand convergence"];
}
}
subgraph cluster_reasoning {
label="Reasoning & Control";
"react" [label="ReAct\n─────────\nThink → Act → Observe\nloop"];
"human-in-the-loop" [label="Human-in-the-Loop\n─────────\nCheckpoint pauses\nfor approval"];
"custom-logic" [label="Custom Logic\n─────────\nCode-driven branching\npattern mixing"];
}
}
| Need | Pattern | Why |
|---|---|---|
| Simple tool-using agent | Single Agent | One model handles everything |
| Fixed-order pipeline | Sequential | Predictable, debuggable flow |
| Independent concurrent tasks | Parallel | Reduce latency via fan-out |
| Repeat until quality met | Loop | Iterative convergence |
| Output needs review gate | Review & Critique | Generator + critic separation |
| Progressive improvement | Iterative Refinement | Multiple drafts toward quality |
| Dynamic task routing | Coordinator | AI decides which specialist |
| Complex decomposition | Hierarchical | Multi-level delegation |
| Debate & convergence | Swarm | Cross-pollination of perspectives |
| Adaptive reasoning | ReAct | Think-act-observe flexibility |
| Critical decision points | Human-in-the-Loop | Human oversight at checkpoints |
| Complex branching logic | Custom Logic | Maximum orchestration flexibility |
digraph builder_flow {
rankdir=TB;
"Identify Pattern" [shape=box];
"Pattern clear?" [shape=diamond];
"Use Selection Guide" [shape=box];
"Load Pattern Builder" [shape=box];
"Validate Inputs" [shape=box];
"Design Components" [shape=box];
"Wire Orchestration" [shape=box];
"Verify" [shape=box];
"Done" [shape=doublecircle];
"Identify Pattern" -> "Pattern clear?";
"Pattern clear?" -> "Load Pattern Builder" [label="yes"];
"Pattern clear?" -> "Use Selection Guide" [label="no"];
"Use Selection Guide" -> "Load Pattern Builder";
"Load Pattern Builder" -> "Validate Inputs";
"Validate Inputs" -> "Design Components";
"Design Components" -> "Wire Orchestration";
"Wire Orchestration" -> "Verify";
"Verify" -> "Done";
}
If the user names a pattern, load its builder file directly. If not, use the Selection Guide table above to recommend one based on their requirements.
Read the corresponding builder file from this skill's directory:
| Pattern | File |
|---|---|
| Single Agent | single-agent.md |
| Sequential | sequential.md |
| Parallel | parallel.md |
| Loop | loop.md |
| Review & Critique | review-critique.md |
| Iterative Refinement | iterative-refinement.md |
| Coordinator | coordinator.md |
| Hierarchical | hierarchical.md |
| Swarm | swarm.md |
| ReAct | react.md |
| Human-in-the-Loop | human-in-the-loop.md |
| Custom Logic | custom-logic.md |
Read the file from the same directory as this SKILL.md, then follow its builder instructions.
Each builder file contains:
Run the smoke test for the pattern to verify the implementation works:
smoke-agentic-patterns with the pattern name as argumentProvides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
npx claudepluginhub bflandev/threadline-ai