F2 · Planning
Poor Task Decomposition
Detects when task breakdown creates subtasks that are impossible, circular, vague, too granular, or too broad. Critical for complex multi-step agent workflows.
Examples
- Task decomposed into subtasks with circular dependencies (A needs B, B needs A)
- Subtask says "handle the infrastructure" with no specifics
- Simple "add button" task over-decomposed into 15 steps when 3 would suffice
- Complex system design has only 2 subtasks, each too broad to execute
Detection methods
- Dependency Analysis
- Detects circular, missing, or impossible dependencies
- Granularity Check
- Validates task-aware decomposition depth (complex vs simple)
- Vagueness Detection
- Flags non-actionable steps using indicator words
- Complexity Estimation
- Identifies subtasks too broad for single execution
Calibration accuracy
F1
0.727
Precision
0.727
Recall
0.727
From the Pisama calibration set. See detector scoreboard for the full table.
Subtypes
- impossible subtask
- missing dependency
- circular dependency
- duplicate work
- wrong granularity
- vague subtask
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.