Skip to content

CHANGELOG.md

Latest commit

 

History

History
102 lines (76 loc) · 3.74 KB

CHANGELOG.md

File metadata and controls

102 lines (76 loc) · 3.74 KB
Raw

Changelog

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

[0.6.3] - 2026-02-17

Added

  • /tinman oilcan command for plain-language Oilcan setup and stream status.
  • /tinman oilcan --json for machine-readable status output.

Changed

  • Update skill metadata/description to include Oilcan event streaming and setup helper.
  • Add feedback/updates contact reference to @cantshutup_.

[0.6.2] - 2026-02-08

Added

  • Local live event stream for dashboards/visualizers: emits ~/.openclaw/workspace/tinman-events.jsonl.
  • /tinman scan, /tinman sweep, and /tinman watch now append structured JSONL events (best-effort, never breaks skill execution).

Changed

  • watch now blocks non-loopback Gateway URLs by default; remote endpoints require --allow-remote-gateway.
  • JSONL event emission now applies basic redaction/truncation for obvious secret-like values.

[0.6.1] - 2026-02-08

Changed

  • Update eval dependency to tinman-openclaw-eval>=0.3.2 (288 probes, updated categories/aliases).
  • Update skill metadata and docs to reflect current probe count and category names.
  • Remove remaining non-ASCII warning glyphs to avoid encoding issues in terminals/Clawdhub packaging.

[0.6.0] - 2026-02-01

Added

  • Security Check System - /tinman check <tool> <args> for pre-execution security verification
  • Security Modes - /tinman mode <level> with three protection levels:
    Mode SAFE REVIEW (S1-S2) BLOCKED (S3-S4)
    safer (default) Proceed Ask human Block
    risky Proceed Auto-approve Block
    yolo Proceed Auto-approve Warn only
  • Allowlist Management - /tinman allow and /tinman allowlist commands
  • Verdict System - SAFE/REVIEW/BLOCKED with severity, confidence, and recommendations
  • 168 Detection Patterns across 16 categories:
    • credential_theft (39), windows_attack (32), crypto_wallet (15)
    • linux_persistence (12), network_exfil (11), macos_attack (9)
    • shell_injection (7), browser_data (6), destructive (6)
    • privilege_escalation (5), mcp_attack (5), git_hooks (5)
    • cloud_metadata (4), container_escape (4), process_spawn (4), evasion (4)
  • Agent Self-Protection - Add to SOUL.md for autonomous security enforcement

Changed

  • Refactored suspicious tool detection to use new CheckResult system
  • Consolidated all legacy patterns into categorized PATTERN_CATEGORIES

[0.5.1] - 2026-02-01

Added

  • /tinman init command for easy workspace setup

Fixed

  • Watch stop now works with PID-based process management
  • Fixed crash when tool args are dict instead of string

[0.5.0] - 2026-02-01

Added

  • Evasion/Bypass detection (30 probes): Unicode homoglyphs, URL/base64/hex encoding, shell injection
  • Memory poisoning attacks (25 probes): Context injection, RAG poisoning, history fabrication
  • Platform-specific attacks (35 probes): Windows, macOS, Linux, cloud metadata
  • 50+ new suspicious tool detection patterns

Changed

  • Total probes increased from 180 to 270+
  • Attack categories expanded from 10 to 13

[0.4.0] - 2026-01-31

Added

  • Financial/Crypto attacks (26 probes): BTC, ETH, SOL, Base wallet theft
  • Unauthorized action detection (28 probes)
  • MCP server attacks (20 probes)
  • Indirect injection (20 probes)

[0.3.0] - 2026-01-30

Added

  • Real-time monitoring via Gateway WebSocket
  • /tinman watch command with realtime and polling modes

[0.2.0] - 2026-01-29

Added

  • /tinman sweep - Proactive security testing with synthetic attack probes

[0.1.0] - 2026-01-28

Added

  • Initial release
  • /tinman scan - Analyze recent sessions
  • /tinman report - View findings
  • Core failure classification for prompt injection, tool exfil, context bleed