F5 · Planning
Flawed Workflow Design
Detects structural problems in agent workflow graphs including unreachable nodes, dead ends, missing error handling, bottlenecks, and missing termination conditions.
Examples
- Workflow has a node that can never be reached from the start node
- Agent graph has a path with no terminal node -- workflow never ends
- AI processing nodes have no error handling -- single failure crashes entire workflow
- All paths funnel through a single bottleneck node creating a scalability issue
Detection methods
- Graph Traversal
- Checks reachability of all nodes from start
- Dead End Detection
- Identifies paths with no terminal nodes
- Error Handler Audit
- Verifies error handling on critical nodes
- Bottleneck Analysis
- Detects nodes with disproportionate in-degree
Calibration accuracy
F1
0.797
Precision
0.851
Recall
0.750
From the Pisama calibration set. See detector scoreboard for the full table.
Subtypes
- unreachable node
- dead end
- missing error handling
- bottleneck
- missing termination
Detect this in production with the framework adapters (LangGraph, CrewAI, AutoGen, OpenAI Agents SDK, Claude Agent SDK, n8n, Dify). See the full taxonomy at /taxonomy.