Getting Started

UT Agent

UT Agent is responsible for unit testing tasks, supports Python/C/C++/C#/Java multi-language, coverage analysis, WinAMS integration, CFG analysis, etc.

---

v3.0.6 (Latest)

Release Date: August 14, 2026 | Base Version: v3.0.5

1. Overview

Item	Content
Release Date	August 14, 2026
Version	v3.0.6
Base Version	v3.0.5

2. Release Notes

2.1 New Features

Module	Details
SWE.4 Diagnostic Pipeline	Added a standalone SWE.4 diagnostic entry that chains intake, source code analysis, test design, specification generation, execution, result review, and report generation.
Unit Test Preparation	Supports source code/AST analysis, function IO and MC/DC extraction, stub function generation, and analysis artifact validation.
Host Diagnostic Execution	Supports Host fixture generation, stub call count and global variable binding, test execution, and coverage report output.

2.4 Bug Fixes

Module	Details
Specification and Result Review	Improved WinAMS Excel/CSV specification generation and diagnostic result review; the minimum, maximum, and representative values of stub function call counts are uniformly displayed as "-" per the rules.

2.5 Known Issues

Module	Details
Qualification Boundary	Host results are for diagnostic purposes only and do not constitute WinAMS validation or formal SWE.4 engineering evidence.

---

v3.0.5

Release Date: August 7, 2026 | Base Version: v3.0.4

1. Overview

Item	Content
Release Date	August 7, 2026
Version	v3.0.5
Base Version	v3.0.4

2. Release Notes

2.1 New Features

Module	Details
WinAMS Skill-Agent Case Generation	1. Upgraded embedded WinAMS case generation from built-in Pipeline Prompt/LLM to a Skill-Agent collaboration mode. 2. Added a phased workflow of prepare → generate testcase.raw.json → validation and repair → finalize. 3. Added task_package, raw testcase Schema, IO whitelist, and MCDC path constraints.
Boundary and Batch Case Capabilities	1. Added automatic boundary case generation and interval solving, supporting all-min, all-max, representative value scanning, and pointer NULL scenarios. 2. Supports multi-target task packages, task sharding, partial submission, and all-or-none gating.
C Case Run Coverage Capability	Added embedded C function case execution and coverage prompting capabilities, supporting pointer size validation, NULL input, and non-branch output write handling.

2.2 Changes

Module	Details
WinAMS Pipeline Refactoring	1. Adjusted the case phase from internal LLM generation to deterministic validation and persistence of Agent raw results. 2. Refactored WinAMS bridging, phase contracts, execution state, and artifact management.
AST and IO Standardization	1. Unified AST parsing using Python/libclang, enhanced handling of text encoding, switch/case, function pointers, and complex member access. 2. Unified integer type min/max normalization and IO conversion rules.
Artifact and Documentation Optimization	Optimized artifact file name generation, safe cleanup, and custom prefixes; supplemented documentation on the WinAMS dedicated pipeline and old/new case generation schemes.

2.3 Removed

Module	Details
Legacy LLM and Redundant Components	Removed the Python-side built-in LLM case generation, LangChain local adaptation layer, legacy GUI, standalone IO extraction step, and redundant log/instruction files.

2.4 Bug Fixes

Module	Details
Code Analysis Fixes	1. Fixed loss of case/default extraction in switch and nested fallthrough handling issues. 2. Fixed issues with function pointer arrays, member function pointers, and Stub target extraction.
Pipeline and Contract Fixes	Fixed abnormal Pipeline property dependency resolution, improved raw Schema, task packages, path reachability, and result post-processing contracts.
Embedded Execution Fixes	Fixed C Case Run pointer parameter validation order, non-decimal size parsing, zero-length pointer NULL passing, and coverage mapping issues.

2.5 Known Issues

Module	Details
Continuous Improvement Items	1. The semantic accuracy of expected values for complex struct, pointer, and array scenarios still requires ongoing validation. 2. End-to-end compatibility across different embedded compilation environments and real WinAMS projects still needs expanded validation coverage.

---

v3.0.4

Release Date: 2026-07-08 | Base Version: v3.0.3

1. Overview

Item	Content
Release Date	2026-07-08
Version	v3.0.4
Base Version	v3.0.3

2. Release Notes

2.1 New Features

Module	Details
WinAMS 6-Step Pipeline	1. Implemented complete WinAMS 6-step pipeline architecture: entry routing → preprocessing → AST parsing → stub function generation → test case generation → batch execution 2. Changed AST parsing code from exe-based to Python + libclang implementation, supports expected value auto-completion 3. Added AST analyzer service and expected value generation service 4. Supports batch mode entry determination and optional output directory parameter
AST Parsing Optimization	1. Merged libclang parsing of same-source files into one pass, downstream reads shared cache, eliminates redundant parsing overhead 2. AST code migration and ast_to_python conversion layer refactoring, fixed IO recognition issues 3. Enhanced preprocessing compatibility: environment recognition, parallel processing, branch recognition library replaced with libclang 4. Execution steps refactored into three stages: "preprocessing → parsing → stub functions"
Build Tool Environment	Refactored build tool environment configuration and dependency management, improved environment switching consistency and maintainability

2.2 Changes

Module	Details
Pipeline Refactoring	Refactored code structure and implemented UT cache acceleration, fixed branch extraction empty issue
WinAMS Refactoring	1. Removed hardcoded libclang.dll path, changed to dynamic lookup 2. Fixed AST analyzer module import path 3. Optimized worksheet selection logic and extracted common functions 4. Migrated documentation to external, added dependency libraries to requirements.txt
Documentation Updates	1. Updated SKILL.md documentation content, removed log cache 2. Added C language data type support matrix documentation 3. Added UT-Agent efficiency comparison documentation

2.3 Removed

Module	Details
None	None

2.4 Bug Fixes

Module	Details
AST Parsing Fixes	1. Fixed AST analyzer module import path 2. Fixed ast_to_python IO recognition issue
Dependency & Build Fixes	1. Added libclang dependency information to requirements.txt 2. Updated .gitignore configuration 3. Deleted invalid files, cleaned up log cache
Pipeline Fixes	1. Fixed branch extraction empty issue 2. Modified preprocessing compatibility, reduced processing steps and implemented parallel processing

2.5 Known Issues

Module	Details
WinAMS Improvements	1. Expected value generation coverage for complex data types (structs/pointers/arrays) still needs validation and improvement. 2. The exception branch of WinAMS batch mode (full rollback when a single test case fails) is not yet closed-loop. 3. Hardware mock solutions for embedded targets other than WinAMS (e.g., ARM Cortex-M) are still under investigation.

---

v3.0.3

Release Date: 2026-06-12 | Base Version: v3.0.2

1. Overview

Item	Content
Release Date	2026-06-12
Version	v3.0.3
Base Version	v3.0.2

2. Release Notes

2.1 New Features

Module	Details
WinAMS Integration Support	1. Added WinAMS dedicated integration (winams-stubs/winams-testcases), supports mock and test case generation 2. Added tool path resolution function and set defaults, resolved blocking points, output directory parameter is optional
C# CFG Adapter	Added C# CFG adapter, supports statement-level and expression-level branch extraction, extends multi-language analysis capabilities
Coverage Closure Enhancement	Added workflow-level stop conditions and module/package target scope support to coverage closure
Test Case Complement	Supplemented complete functional test cases, C# closure and routing regression tests, added core capabilities for test design and path analysis

2.2 Changes

Module	Details
Schema Refactoring	Refactored JSON schema and enhanced test_run and contract tools, unified artifact metadata fields
WinAMS Refactoring	Adjusted WinAMS module import paths and dependency configuration, replaced logger with print function and optimized code format
Documentation Fixes	Fixed inconsistent parts in md, updated skill description documentation
Artifact Validation & Recommendations	Artifacts uniformly carry schema_version/trace_id/created_at metadata, added recommendation_cards layered output
Documentation Sync	Synchronized aiko skill usage instructions and workflow documentation, updated skill descriptions

2.3 Removed

Module	Details
None	None

2.4 Bug Fixes

Module	Details
Import & Hardcode Fixes	Fixed missing imports, eliminated hardcoded default language, completed contracts and schema
Context Fixes	Cleaned up redundant fields in context_build payload, isolated tracking metadata. Enhanced context dependency classification and Python call extraction
Coverage Fixes	Fixed stability issues in context building and coverage aggregation
CFG Analyzer Fixes	Changed cfg_analyzer adapter registration to standard method, removed global dictionary and side effects
C++ Parsing Fixes	Fixed end_line recognition error for C++ out-of-class methods, enhanced test file judgment logic
Skill Routing Fixes	Added WinAMS dedicated integration support, optimized routing configuration to ensure correct routing of WinAMS requests
Coverage Closure Fixes	Completed iteration rules, fixed workflow field ownership, cleaned up c_cpp hardcode residues. Covers New Java closure and contract and other optimizations

---

v3.0.2

Release Date: 2026-05-29 | Base Version: v3.0.1

1. Overview

Item	Content
Release Date	2026-05-29
Version	v3.0.2
Base Version	v3.0.1

2. Release Notes

2.1 New Features

Module	Details
Test Design + CFG Analysis	Added core capabilities for test design and path analysis
Python Support + Multi-language Routing	Completed Python support and improved multi-language framework routing

2.2 Changes

Module	Details
Documentation Fixes	Fixed inconsistent parts in md
Redundant Features	Removed unnecessary MCP and intermediate files

2.3 Removed

Module	Details
None	None

2.4 Bug Fixes

Module	Details
API & Experience Optimization	1. Python language edge scenario support still needs further improvement 2. MCP subAgent tool call experience and coverage gap intelligent analysis still have room for optimization

---

v3.0.1

Release Date: 2026-05-22 | Base Version: v3.0.0

1. Overview

Item	Content
Release Date	2026-05-22
Version	v3.0.1
Base Version	v3.0.0

2. Release Notes

2.1 New Features

Module	Details
Python Language Support	1. Added Python project analysis, test execution and coverage parsing capabilities 2. Implemented source code parsing based on AST, no additional parsing dependencies required 3. Automatically identifies pytest, unittest, coverage.py related configurations and commands
MCP SubAgent Adaptation Layer	1. Added UT-Agent MCP subAgent adaptation layer, provides manifest.yaml, agent.md, plugin.py 2. Provides 4 scenario tools and 6 diagnostic tools
Coverage Improvement Scenario Enhancement	1. plan_ut_execution supports targets batch planning 2. analyze_ut_target added recommended_test_order and reusable_fixtures output
Gradle Support	Added Gradle build system support, extends build tool coverage for Java projects
Technical Optimization	Completed Phase 2 Service API refactoring, first four modules implement Service API direct connection, extracted module service interfaces such as command_detect, coverage_analyze, test_run

2.2 Changes

Module	Details
Connection & Intent Generation	1. Fixed SubAgent connection issue 2. Improved MCP test intent generation target inference and class-level branch output
Other Fixes	Fixed ut-agent skill related issues, and fixed multiple known defects

2.3 Removed

Module	Details
None	None

2.4 Bug Fixes

Module	Details
API & Experience Optimization	1. Python language edge scenario support still needs further improvement 2. MCP subAgent tool call experience and coverage gap intelligent analysis still have room for optimization