Hypothesis Debugger Investigator

Overview

This agent acts as a structured, hypothesis-driven debugger. Instead of randomly searching the codebase, it forces the investigation around one specific theory about why a bug is occurring. It follows a rigorous protocol to gather concrete evidence necessary to confirm or falsify the initial assumption.

Capabilities

• Hypothesis Parsing: Breaks down complex failure theories into testable components.
• Evidence Definition: Clearly defines what evidence would prove (confirm) or disprove (falsify) the hypothesis.
• Multi-Source Gathering: Searches across source code, logs, and potential execution paths for supporting data.
• Confidence Scoring: Provides a structured assessment of certainty based on gathered evidence, citing specific file:line locations.
• Structured Reporting: Delivers a final report detailing the causal chain, confirming evidence, and any remaining gaps.

Example Use Cases

• Investigating Race Conditions: Given the hypothesis that 'Thread A overwrites data written by Thread B due to improper locking,' this agent will trace all shared resource access points.
• Analyzing Performance Degradation: If the theory is that 'Database query X times out because of an unindexed join,' the agent will focus its evidence gathering on the ORM layer and database schema definitions.
• Debugging State Management Issues: When suspecting a specific module's state variable is being incorrectly mutated, the agent will trace all read/write operations for that variable across the application lifecycle.