Quality phase04: working-method prompts, compose demotion (Backend #121) #290

2026-10-09T23:44:00Z

weichen commented

2026-10-09 23:44:00 +00:00

摘要

P4 收口 PR（组件 Backend #121 已合入）。Dev 任务卡 feature/agent-quality-phase04；主记录 §4 P4（裁定 4）。

base.method 工作方法段 + 工具描述重写 + 计划提醒常开 + composeFinal 默认关（空正文兜底）
任务卡：make test + task-sync + 真实模型对照 evidence；before/after 报告 .task/phase04-eval/after-p4.json

真实模型对照（before=P3 后 main）

指标	P3 后	P4 后
plan-followup 工具总数（碎片螺旋）	117	47
计划使用（followup/pm-tool）	0/0	8/5
检查失败	4	0

（组合 make test 首跑遇 SSE 时序 flake——单测/复跑均绿——终跑绿行已入卡。）

剩余

P5（评分语义死循环清理、事实一致性核查——未核实数字仍 10、异模型验证器）与模型 A/B（无预算上限）。

## 摘要 P4 收口 PR（组件 Backend #121 已合入）。Dev 任务卡 feature/agent-quality-phase04；主记录 §4 P4（裁定 4）。 - base.method 工作方法段 + 工具描述重写 + 计划提醒常开 + composeFinal 默认关（空正文兜底） - 任务卡：make test + task-sync + 真实模型对照 evidence；before/after 报告 `.task/phase04-eval/after-p4.json` ## 真实模型对照（before=P3 后 main） | 指标 | P3 后 | **P4 后** | |---|---|---| | plan-followup 工具总数（碎片螺旋） | 117 | **47** | | 计划使用（followup/pm-tool） | 0/0 | **8/5** | | 检查失败 | 4 | **0** | （组合 make test 首跑遇 SSE 时序 flake——单测/复跑均绿——终跑绿行已入卡。） ## 剩余 P5（评分语义死循环清理、事实一致性核查——未核实数字仍 10、异模型验证器）与模型 A/B（无预算上限）。

weichen added 9 commits 2026-10-09 23:44:01 +00:00

Record the task card for feature-agent-quality-phase04 2cbb00befa

Why: establish the authorized task workspace and its initial baseline.

Evidence: preserved precheck receipt in .task/feature-agent-quality-phase04.md; R0=4713f04c94eb2512152e412fd010f846caa10021.

Impact: task metadata only; the feature workspace starts at 4713f04c94.

Refs: .task/feature-agent-quality-phase04.md

Point the backend gitlink at the phase04 head 8b09cdc411

Why: The Dev combination must test the working-method prompt set and the
compose demotion.

Evidence: SELF - gitlink equals backend feature/agent-quality-phase04 head.

Impact: Pointer only.

Record the sync checkpoint 27246370b7

Why: bind the task validation checkpoint to the latest integration baseline.

Evidence: SELF recursive component checks; sync_base=4713f04c94eb2512152e412fd010f846caa10021; candidate=8b09cdc411fa59b8a53a82c746c7ba42913574a0.

Impact: task metadata only.

Refs: .task/feature-agent-quality-phase04.md

Record the phase04 make-test and sync evidence dd18b9fd6f

Record the phase04 live eval and result summary 80a35b888b

Point the backend gitlink at the merged phase04 head d5185c5ff3

Record the sync checkpoint c14193353f

Why: bind the task validation checkpoint to the latest integration baseline.

Evidence: SELF recursive component checks; sync_base=4713f04c94eb2512152e412fd010f846caa10021; candidate=d5185c5ff32d909a823237b5c264b1e3007c3476.

Impact: task metadata only.

Refs: .task/feature-agent-quality-phase04.md

Close out phase04 with merge links and final evidence c03ddc2652

Record the phase04 final green make-test evidence (SSE flake rerun)

ci / test (pull_request) Successful in 2m12s

Details

d78f3bedf5

weichen merged commit b7489d4431 into main

2026-10-09 23:48:28 +00:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: weichen/Agent-System-Dev#290