AgentSkillsCN

detecting-memory-leaks

通过识别未释放的内存分配和错误路径中缺失的清理，检测内存泄漏漏洞。在分析长时间运行的进程、资源管理或调查内存耗尽问题时使用。

SKILL.md
--- frontmatter
name: detecting-memory-leaks
description: Detects memory leak vulnerabilities by identifying unfreed memory allocations and missing cleanup in error paths. Use when analyzing long-running processes, resource management, or investigating memory exhaustion issues.

Memory Leak Detection

Detection Workflow

  1. •Identify allocation points: Find all malloc/calloc/realloc calls, new/delete pairs (C++), map allocation locations
  2. •Trace allocation lifecycle: Use xrefs_to to trace allocations, follow pointers through code, identify all free() calls
  3. •Check cleanup paths: Verify every allocation has a corresponding free, check all error handling paths, review exception handling, assess callback cleanup
  4. •Assess impact: How much memory is leaked? How often does it leak? What's the cumulative impact? Can it cause DoS?

Key Patterns

  • •Missing free: malloc() without corresponding free(), calloc() without free(), new without delete (C++)
  • •Exception path leaks: allocations before exception-prone operations, missing cleanup in error handling, leaks in early returns
  • •Reference cycles: circular references in data structures, reference counting bugs, shared pointer cycles
  • •Long-lived allocations: allocations that never get freed, caches without size limits, accumulating data structures

Output Format

Report with: id, type, subtype, severity, confidence, location, allocation (function, address, size), leak path, leak scenario, leak size per call, frequency, exploitability, impact, mitigation.

Severity Guidelines

  • •HIGH: Large leaks in frequently called functions
  • •MEDIUM: Small leaks or infrequent allocations
  • •LOW: Minor leaks with limited impact

See Also

  • •patterns.md - Detailed detection patterns and exploitation scenarios
  • •examples.md - Example analysis cases and code samples
  • •references.md - CWE references and mitigation strategies