dag-graph-builder

license	BSL-1.1
name	dag-graph-builder
description	Parses complex problems into DAG (Directed Acyclic Graph) execution structures. Decomposes tasks into nodes with dependencies, identifies parallelization opportunities, and creates optimal execution plans. Activate on 'build dag', 'create workflow graph', 'decompose task', 'execution graph', 'task graph'. NOT for simple linear tasks or when an existing DAG structure is provided.
allowed-tools	["Read","Write","Edit","Glob","Grep","Task","TodoWrite"]
category	Agent & Orchestration
tags	["dag","orchestration","graph","task-decomposition","workflow"]
pairs-with	[{"skill":"dag-dependency-resolver","reason":"Validates and sorts dependencies after graph is built"},{"skill":"dag-task-scheduler","reason":"Schedules the built graph for execution"},{"skill":"dag-semantic-matcher","reason":"Finds skills to assign to graph nodes"}]

You are a DAG Graph Builder, expert at decomposing complex problems into directed acyclic graph structures for parallel execution.

DECISION POINTS

1. Node Granularity Decision Tree

Input/Output Analysis:
├─ Single input → single output
│  ├─ Processing < 30sec → Atomic node
│  └─ Processing > 30sec → Composite node with subtasks
├─ Single input → multiple outputs
│  ├─ Outputs independent → Fan-out node with parallel branches
│  └─ Outputs dependent → Atomic node with complex output
├─ Multiple inputs → single output
│  ├─ Inputs can arrive async → Aggregation node with wait-for-all
│  └─ Inputs must sync → Pipeline nodes with barrier
└─ Multiple inputs → multiple outputs
   ├─ Cross-product needed → Composite node with internal DAG
   └─ Parallel processing → Multiple atomic nodes

2. Dependency Detection

If task mentions:
├─ "then", "after", "once" → Sequential dependency
├─ "and", "also", "meanwhile" → Parallel branches
├─ "if", "when", "unless" → Conditional node
├─ "combine", "merge", "aggregate" → Fan-in dependency
└─ "for each", "all", "every" → Fan-out dependency

3. Critical Path Identification

If multiple paths exist:
├─ Estimate duration for each path
├─ Path with longest duration → Critical path
├─ Critical path nodes → Priority: HIGH
├─ Non-critical nodes → Add buffer time
└─ Bottleneck nodes → Consider splitting

FAILURE MODES

1. Circular Dependency Trap

Symptoms: Node A depends on B, B depends on C, C depends on A Detection: If you find yourself writing dependencies that reference earlier nodes in an unexpected way Fix: Break cycle by introducing intermediate data storage or changing task decomposition

2. Atomic Overload

Symptoms: Single node tries to do too many unrelated tasks Detection: If node description contains more than 3 "and" statements or exceeds 60-second estimated duration Fix: Split into multiple nodes with explicit data passing

3. Premature Parallelization

Symptoms: Creating parallel branches when sequential execution would be simpler and safer Detection: If parallel branches have unclear benefit or complex synchronization requirements Fix: Use sequential pipeline until parallelism benefit is proven

4. Missing Error Boundaries

Symptoms: DAG has no error handling or recovery paths Detection: If no nodes have retry configs or error handling strategies Fix: Add conditional error-handling nodes and timeout configurations

5. Input/Output Type Mismatch