Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Update: re-run after edits; refreshed diagnostic output and report formatting.
What This Skill Does
Diagnoses active Cognitive Bullwhip Effects in AI agent systems. Takes a snapshot of recent agent decision history, calculates variance amplification ratios across layers, identifies where a small input error began compounding, and recommends which intervention skill addresses the root cause.
Rather than debugging visible output symptoms, it maps the full amplification chain back to the origin layer so the fix targets the cause, not wherever the failure last surfaced.
When to use it
- Tracing why an agent returns wildly different outputs for identical inputs
- Pinpointing the origin layer of a cascading failure in a multi-step agent pipeline
- Running a pre-deployment diagnostic on a new agent before production rollout
- Identifying whether reasoning drift or input noise is driving erratic agent behavior
- Periodic health checks on long-running agents to catch amplification before it compounds
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: My agent is producing contradictory analyses on the same market data across runs — what's wrong?
- 1Accept decision_log entries with timestamps, input summaries, outcomes, and variance scores
- 2Scan each layer for output variance exceeding input variance by more than 3x
- 3Trace amplification back to the earliest layer where the ratio first spiked
- 4Classify the pattern type based on the origin layer (reasoning layer → reasoning_drift)
- 5Return severity score, amplification map, and recommended intervention skill with urgency level
Bullwhip active at severity 72/100. Origin: reasoning layer, 8.2x amplification ratio. Pattern: reasoning_drift. Recommended action: apply LogicStack immediately.