Skip to content

CHANGELOG.md

Latest commit

 

History

History
80 lines (61 loc) · 3.16 KB

CHANGELOG.md

File metadata and controls

80 lines (61 loc) · 3.16 KB

Changelog

All notable changes to ForceField will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.7.2] - 2026-03-27

Added

  • forcefield init -- scaffold .forcefield/constitution.yaml for vibe coding governance (default/strict/permissive templates)
  • guard.audit_report() -- generate structured JSON or Markdown audit reports from scan events
  • guard.eval() -- run security eval suites (116 built-in attacks or custom YAML)
  • Constitution engine -- YAML-driven governance rules for files, commands, tools, and content
  • guard.scan_command() -- scan terminal commands for 22 dangerous patterns
  • guard.scan_filename() -- scan filenames for 12 security-sensitive patterns
  • guard.protect_path() / guard.is_protected() -- glob-based protected path management
  • CLI commands: forcefield init, forcefield eval, forcefield scan-command, forcefield scan-filename
  • GitHub Action for CI/CD integration (GitHub Marketplace listed)
  • pre-commit hook support
  • Homebrew tap (brew tap datasciencetech/forcefield)
  • npm wrapper (npx forcefield-ai)
  • VS Code extension with Sentinel Mode
  • Open VSX Registry listing
  • JetBrains Marketplace plugin

Changed

  • Multi-turn session tracker now detects crescendo and probe-then-inject sequences
  • Anti-obfuscation normalizer handles zero-width chars, homoglyphs, leetspeak, and mixed scripts

[0.6.0] - 2026-03-15

Added

  • ONNX ML model for prompt injection detection (95%+ accuracy, 235KB)
  • TF-IDF + Random Forest ensemble scoring
  • Endpoint security testing (forcefield test <url>)
  • Cloud hybrid scoring via ForceField Gateway API
  • Output moderation (hate speech, violence, credential leaks)
  • Chat template validation (forcefield validate-template)
  • LangChain callback handler (langchain-forcefield package)
  • LlamaIndex callback handler (llama-index-forcefield package)
  • OpenAI drop-in wrapper (ForceFieldOpenAI)
  • FastAPI middleware (ForceFieldMiddleware)

[0.5.0] - 2026-02-28

Added

  • PII detection for 18 types (SSN, credit card, email, phone, IBAN, JWT, API keys, etc.)
  • PII redaction with configurable replacements
  • Jinja2 chat template backdoor scanning
  • Token smuggling and payload splitting detection
  • Source file auditing (forcefield audit <path>)
  • JSON and Markdown report output formats

[0.4.0] - 2026-02-15

Added

  • Prompt injection detection across 12 categories with 60+ regex patterns
  • Jailbreak and role escalation detection (DAN-style attacks)
  • Data exfiltration detection (obfuscated destinations, JSON payloads)
  • System prompt extraction detection
  • Sensitivity levels (low/medium/high/critical) with configurable thresholds
  • CLI selftest with 116 built-in attack prompts
  • HTTP proxy mode (forcefield serve)

[0.3.0] - 2026-02-01

Added

  • Initial public release
  • Core Guard class with scan(), redact(), moderate() methods
  • Regex-based detection engine (81% detection rate)
  • CLI: forcefield scan, forcefield redact, forcefield selftest
  • Python 3.9-3.13 support
  • Zero external dependencies for core package