AX Agent claw-code 진척율 문서화 및 메시지 축 단순화
Some checks failed
Release Gate / gate (push) Has been cancelled
Some checks failed
Release Gate / gate (push) Has been cancelled
- claw-code 대비 AX Agent 진척율을 README, DEVELOPMENT, parity 계획 문서에 수치로 기록 - 핵심 엔진 영향 설정 최소화 원칙을 정리하고 계획 모드 잔재 노출을 추가로 축소 - ChatWindow 메시지 버블과 좌측 대화 목록 카드를 더 얇은 밀도로 조정해 claw-code식 읽기 축에 가깝게 정리 - 검증: dotnet build src/AxCopilot/AxCopilot.csproj -c Release -v minimal -p:OutputPath=bin\\verify\\ -p:IntermediateOutputPath=obj\\verify\\ (경고 0 / 오류 0)
This commit is contained in:
@@ -1,5 +1,8 @@
|
||||
# AX Copilot - 媛쒕컻 臾몄꽌
|
||||
|
||||
- Document update: 2026-04-05 16:55 (KST) - Recorded the current `claw-code` parity estimate for AX Agent: core execution engine `82%`, main chat UI `68%`, Cowork/Code status UX `63%`, internal settings linkage `88%`, overall AX Agent parity `74%`.
|
||||
- Document update: 2026-04-05 16:55 (KST) - Added an engine-settings review rule for ongoing cleanup: settings that materially alter the main execution route should be minimized, kept developer-only when necessary, or removed from user-facing surfaces when they no longer represent real runtime choices. Plan-mode remnants were reduced further as part of this pass.
|
||||
- Document update: 2026-04-05 16:55 (KST) - Simplified AX Agent message rows and sidebar conversation items toward the `claw-code` reading model. Message bubbles now use tighter padding/radius/meta text, and conversation rows now prefer lightweight running/failure summary text over heavier success/failure badge cards.
|
||||
- Document update: 2026-04-05 16:02 (KST) - Fixed the Cowork document execution handoff around `document_plan`. The loop no longer depends on broken localized marker strings to detect the scaffold/body block or the immediate-next-step hint; it now extracts body markers robustly and resolves the correct follow-up tool (`html_create`, `document_assemble`, `docx_create`, `markdown_create`) before re-prompting the model.
|
||||
- Document update: 2026-04-05 16:02 (KST) - Added `ResolveEffectivePlanMode(...)` so Cowork document/content tasks automatically use the `always` plan path even when the persisted plan mode is `off`. This brings Cowork closer to the `claw-code` expectation of plan-first execution for document-heavy work.
|
||||
- Document update: 2026-04-05 16:02 (KST) - Strengthened `BuildCoworkSystemPrompt()` so document/report/proposal/manual requests must produce an execution plan first and are not considered complete until a real output file path has been created or updated.
|
||||
|
||||
Reference in New Issue
Block a user