AgentSkillsCN

debugging

系统化调试策略与技巧

SKILL.md
--- frontmatter
name: debugging
description: Systematic debugging strategies and techniques

Debugging Skill

Objectives

Provide systematic methodology for debugging code:

  • •Identify bugs efficiently
  • •Find root causes
  • •Fix without side effects
  • •Prevent regression

Debugging Strategies

1. Binary Search Debugging

When bug location is unknown:

  1. •Identify range where bug exists
  2. •Test midpoint
  3. •Narrow range based on result
  4. •Repeat until isolated

2. Trace-Based Debugging

For flow-related bugs:

  1. •Add logging at key points
  2. •Trace execution path
  3. •Compare expected vs actual flow
  4. •Identify divergence point

3. State Inspection

For data-related bugs:

  1. •Capture state at key points
  2. •Compare with expected state
  3. •Identify corruption point
  4. •Trace data source

4. Regression Testing

When something "used to work":

  1. •Identify when it broke (git bisect)
  2. •Compare working vs broken code
  3. •Isolate the change
  4. •Understand why change broke it

Bug Categories

Logic Errors

  • •Off-by-one errors
  • •Incorrect conditionals
  • •Wrong operator precedence
  • •Missing edge cases

Data Errors

  • •Null/undefined references
  • •Type mismatches
  • •Invalid state transitions
  • •Race conditions

Integration Errors

  • •API contract violations
  • •Protocol mismatches
  • •Version incompatibilities
  • •Configuration issues

Performance Bugs

  • •Memory leaks
  • •Inefficient algorithms
  • •Blocking operations
  • •Resource exhaustion

Debugging Tools Usage

Logging

python
# Strategic logging
logger.debug(f"Function entry: {func.__name__}, args={args}")
logger.debug(f"State: {vars(self)}")
logger.debug(f"Exit with result: {result}")

Assertions

python
# Catch invalid states early
assert data is not None, "Data should not be None at this point"
assert len(items) > 0, "Items list should not be empty"

Breakpoints

  • •Set at suspected locations
  • •Conditional breakpoints for specific cases
  • •Watch expressions for key variables

Common Patterns

The Scientific Method

  1. •Observe: Gather data about the bug
  2. •Hypothesize: Form theory about cause
  3. •Predict: What would prove/disprove hypothesis
  4. •Test: Execute debugging action
  5. •Analyze: Evaluate results
  6. •Iterate: Refine hypothesis

Rubber Duck Debugging

  1. •Explain the code line by line
  2. •Describe expected behavior
  3. •Describe actual behavior
  4. •Identify discrepancy

Output Specification

Debugging reports should include:

  • •Clear problem statement
  • •Steps taken to investigate
  • •Findings and evidence
  • •Root cause identification
  • •Proposed fix
  • •Verification approach