cognitive-bullwhip
jkc3080 द्वारा
Coding Agents & IDEsसंक्षिप्त
Diagnoses whether a Cognitive Bullwhip Effect is already active in your agent system.
विवरण
यह स्किल क्या करती है
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.
इसे कब इस्तेमाल करें
- 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
उदाहरण वर्कफ़्लो
User asks: My agent is producing contradictory analyses on the same market data across runs — what's wrong?
- 1
Accept decision_log entries with timestamps, input summaries, outcomes, and variance scores
- 2
Scan each layer for output variance exceeding input variance by more than 3x
- 3
Trace amplification back to the earliest layer where the ratio first spiked
- 4
Classify the pattern type based on the origin layer (reasoning layer → reasoning_drift)
- 5
Return 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.
इंस्टॉलेशन
clawhub install jkc3080/cognitive-bullwhipCLI कमांड चलाएँ, या प्रॉम्प्ट को Claude Code, Cursor, Codex या जो भी कोडिंग एजेंट आप इस्तेमाल करते हैं उसमें पेस्ट करें।