Skip to content

Latest commit

 

History

History
66 lines (49 loc) · 2.77 KB

CHANGELOG.md

File metadata and controls

66 lines (49 loc) · 2.77 KB

Changelog

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

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

[0.3.0-alpha.0] - 2026-02-14

Added

  • Per-feature scanning modes: deterministic, probabilistic, off
  • New config fields: audit_mode, context_injection_mode, outbound_mode, tool_gating_mode, reminder_mode
  • 3 new probabilistic tools: prisma_airs_scan_prompt, prisma_airs_scan_response, prisma_airs_check_tool_safety
  • Mode-aware guard reminder (adapts content to active modes)
  • Config validation: fail_closed=true rejects probabilistic modes
  • Centralized config module (src/config.ts) with resolveMode(), resolveAllModes()
  • 26 new tests (93 total across 5 test files)

Changed

  • Hook registration is now conditional based on resolved mode
  • Guard handler exports buildReminder(), DETERMINISTIC_REMINDER, PROBABILISTIC_REMINDER
  • Tools handler exports shouldBlockTool(), TOOL_BLOCKS, DEFAULT_HIGH_RISK_TOOLS
  • Outbound handler exports maskSensitiveData(), shouldMaskOnly(), buildBlockMessage()
  • Status RPC response now includes modes object

Removed

  • Deprecated boolean config flags: audit_enabled, context_injection_enabled, outbound_scanning_enabled, tool_gating_enabled, reminder_enabled
    • Use *_mode fields instead

[0.1.4] - 2026-02-04

Changed

  • BREAKING: Updated agent tool API for OpenClaw v2026.2.1+ compatibility
    • Changed handler to execute method signature
    • Added _id: string as first parameter
    • Return type now uses OpenClaw tool result format: { content: [{ type: "text", text: ... }] }
  • Added entrypoint field to openclaw.plugin.json

Fixed

  • Fixed tool.execute is not a function error on OpenClaw v2026.2.1+

[0.1.0] - 2025-01-31

Added

  • Initial release as OpenClaw plugin
  • Pure TypeScript implementation with direct AIRS API integration via fetch()
  • Gateway RPC methods: prisma-airs.scan, prisma-airs.status
  • Agent tool: prisma_airs_scan
  • CLI commands: prisma-airs, prisma-airs-scan
  • prisma-airs-guard hook: injects security scanning reminder on agent bootstrap
  • Session tracking: session_id param to group related scans
  • Transaction correlation: tr_id param for prompt/response pairing
  • Metadata support: app_name, app_user, ai_model params
  • Detection categories: prompt injection, DLP (prompt/response), URL filtering
  • Action types: allow, warn, block
  • Severity levels: SAFE, LOW, MEDIUM, HIGH, CRITICAL
  • Plugin config schema for profile_name, app_name, reminder_mode
  • Vitest test suite (22 tests)
  • ESLint 9 + Prettier + TypeScript strict mode
  • Pre-commit hooks via Husky + lint-staged