F10 · Execution
Communication Breakdown
Detects when messages between agents are misunderstood or misinterpreted, causing incorrect downstream behavior.
Examples
- Agent A sends JSON data but Agent B parses it as plain text
- Ambiguous instruction "process the results" interpreted differently by two agents
- Agent receives conflicting instructions from two upstream agents
- Critical information missing from inter-agent message, causing incomplete handoff
Detection methods
- Intent Alignment
- Measures alignment between sender's intent and receiver's interpretation
- Format Compliance
- Checks message format matches expected schema (JSON, list, code)
- Ambiguity Detection
- Flags semantically ambiguous instructions
- Completeness Check
- Verifies all required information is present in messages
Calibration accuracy
F1
0.818
Precision
0.724
Recall
0.940
From the Pisama calibration set. See detector scoreboard for the full table.
Subtypes
- intent mismatch
- format mismatch
- semantic ambiguity
- incomplete information
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.