Changelog

All notable changes to Akira.

[Unreleased]

Added

• Initial release of Akira LLM Security Testing Framework
• Interactive msfconsole-style console
• Attack modules:
  • injection/basic_injection - Prompt injection testing
  • jailbreak/dan_jailbreak - DAN-style jailbreak attacks
  • extraction/system_prompt_leak - System prompt extraction
  • dos/magic_string - Claude magic string DoS
• Target support:
  • OpenAI API
  • Anthropic Claude API
  • HuggingFace Inference API
  • AWS Bedrock
  • AWS SageMaker
  • Generic API (any LLM-powered endpoint)
• SQLite storage for:
  • Attack history persistence
  • Target profiles
  • Prompt cache
  • Response cache
• Fuzzy search for module discovery
• Rust extensions for performance (optional):
  • Payload fuzzing
  • Pattern matching
  • Response analysis
  • Fuzzy string matching
• CLI for non-interactive usage
• mdbook documentation

Security

• All attack modules are for authorized testing only
• API keys stored locally in SQLite database
• No telemetry or data collection

Versioning

This project follows Semantic Versioning:

• MAJOR: Incompatible API changes
• MINOR: New functionality (backwards compatible)
• PATCH: Bug fixes (backwards compatible)

Reporting Issues

Report bugs and feature requests at: https://github.com/yourusername/akira/issues