@loomhelm/multi-task-flow 0.0.0-stage → 0.2.1

package/toolkit/skills/dev-plan/references/replay-cases/negative-legacy-profile-text.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请按 fast-track 规划 legacy profile 文本。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Legacy Profile Text\n\n## Summary\n- 结论：新计划不能继续输出自由文本 model profile。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 固化 profile 语法\n  新计划必须使用紧凑的 `[model, effort]` 形式。\n\n## Non-goals\n- 不改变 legacy recovery 读取能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:review [gpt-5.6-luna, high] => 只读检查变更; model=gpt-5.6-luna; reasoning_effort=high (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝自由文本 profile。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-main-agent-profile.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请按 fast-track 规划 main-agent profile。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Main Agent Profile\n\n## Summary\n- 结论：main-agent 不应重复声明当前会话 profile。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证 main-agent 继承\n  主执行者继承当前会话模型，不在任务行重复声明。\n\n## Non-goals\n- 不改变委派 agent 的 profile 规则。\n\n## Task Orchestration\n- wave: 1\n  - main-agent [gpt-5.6-luna, high] => 执行实现并准备验证 (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝 main-agent 的 inline profile。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-new-delivery-topology.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"走 fast-track。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Topology\n\n## Summary\n- 结论：错误保留旧字段。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- `delivery_topology`: `direct`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证拒绝\n  新计划不得再带旧 routing 字段。\n\n## Non-goals\n- 不执行实现。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 验证合同 (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝旧字段。\n\n## Assumptions/Defaults\n- fixture 仅用于回放：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-new-mixed-top-level-flow.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"走 initiative。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Mixed Flow\n\n## Summary\n- 结论：错误混用两个顶层 flow。\n\n## Flow Decision\n- `flow_type`: `initiative`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- `integration_branch`: `prd/mixed-demo`\n- `initiative_contract_version`: `2`\n- `child_integration_policy`: `ticket-declared`\n- `child_owner_gate_policy`: `evidence-bound`\n- `final_merge_policy`: `owner-explicit`\n- `ticket_plan_policy`: `inherit-approved-plan`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证拒绝\n  Task Orchestration 只能绑定一个顶层 flow。\n\n## Non-goals\n- 不执行实现。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 错误绑定两个 flow (skill: $initiative-flow + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝双 flow。\n\n## Assumptions/Defaults\n- fixture 仅用于回放：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-specialist-review-effort.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请按 fast-track 规划 formal review。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Formal Reviewer Identity\n\n## Summary\n- 结论：actual reviewer profile 只在运行时记录。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- `review_tier`: `Tier 4`\n- `review_rationale`: `跨协议与持久化边界，需要完整 proof。`\n- `runtime_review_strategy`: `candidate ready 后动态创建 required lanes。`\n- `review_profile_recommendation`: `四至六个 lane 建议 Luna medium，host 不支持时解析其他 pair。`\n- 开发基线：`origin/main`\n\n## Goals\n- G1: 固化 formal review 运行时边界。\n  reviewer identity 不能成为计划事实。\n\n## Non-goals\n- N1: 不锁定具体 model。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:record-integrity [gpt-5.6-sol, ultra] => 运行 formal review wave reviewer (skill: $fast-flow)\n\n## Acceptance & Validation\n- A1 [contract/direct] runtime review | evidence=`test:review` | assertions=`fresh reviewer`\n\n## Assumptions/Defaults\n- host profile resolution 已确定\n</proposed_plan>"}]} }

package/toolkit/skills/dev-plan/references/replay-cases/negative-specialist-review-profile.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请按 normal 规划 formal review。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Formal Review Predeclaration\n\n## Summary\n- 结论：formal review 在 readiness 后运行时创建。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- `review_tier`: `Tier 3`\n- `review_rationale`: `共享审查合同发生变化，需要 current-head proof。`\n- `runtime_review_strategy`: `readiness 通过后运行时创建 fresh formal lanes。`\n- `review_profile_recommendation`: `四个 lane 建议使用当前 host 支持的 Luna medium pair。`\n- 开发基线：`origin/main`\n\n## Goals\n- G1: 保持 reviewer 生命周期边界。\n  formal lane 不在计划中预排。\n\n## Non-goals\n- N1: 不固定后续 Repair Batch 的 reviewer 数量。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:policy-semantics [gpt-5.6-luna, xhigh] => 对冻结 current head 执行 formal code review lane (skill: $normal-flow)\n\n## Acceptance & Validation\n- A1 [contract/direct] formal launch | evidence=`test:review` | assertions=`runtime launch only`\n\n## Assumptions/Defaults\n- readiness gate 已确定\n</proposed_plan>"}]} }

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-contradictory-write.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Contradictory Read Only\n\n## Summary\n- 结论：同一任务不能先禁止写入再要求写入。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- 目标 1: 验证否定语句边界\n  checker 必须识别后续的真实写入指令。\n\n## Non-goals\n- 不允许 sub-agent 修改工作区。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:review [gpt-5.6-luna, xhigh] => must not write to the shared workspace; then write the fix to the shared workspace (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝矛盾任务。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-filesystem-write-actions.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Filesystem Write Actions\n\n## Summary\n- 结论：sub-agent 的常见文件系统写入命令都必须被只读门禁拒绝。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 覆盖命令式写入\n  checker 必须识别 patch、remove 和 permission-changing 写入。\n\n## Non-goals\n- 不开放 sub-agent 的任何文件系统写入能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:patch [gpt-5.6-luna, xhigh] => apply a patch in the current worktree (skill: $normal-flow)\n  - sub-agent:remove [gpt-5.6-luna, xhigh] => remove files from the shared workspace (skill: $normal-flow)\n  - sub-agent:permissions [gpt-5.6-luna, xhigh] => chmod the current file (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有命令式文件系统写入。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-generic-english-write.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Generic English Write\n\n## Summary\n- 结论：普通英文写入目标也必须被只读门禁拒绝。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 覆盖普通英文写入句式\n  checker 必须拒绝 repository、files 与 source code 的写入声明。\n\n## Non-goals\n- 不开放 sub-agent 的任何写入能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:repo [gpt-5.6-luna, xhigh] => write to the repository (skill: $normal-flow)\n  - sub-agent:files [gpt-5.6-luna, xhigh] => modify files (skill: $normal-flow)\n  - sub-agent:code [gpt-5.6-luna, xhigh] => write source code (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有普通英文写入目标。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-generic-implementation-write.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Generic Implementation Write\n\n## Summary\n- 结论：泛化 implementation assignment 也属于 sub-agent 写入。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 拦截无具体文件名的实现任务\n  checker 必须拒绝 functionality 与 solution 写入声明。\n\n## Non-goals\n- 不开放 sub-agent 的任何实现权限。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:impl [gpt-5.6-luna, xhigh] => implement functionality (skill: $normal-flow)\n  - sub-agent:fix [gpt-5.6-luna, xhigh] => implement the solution (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须报告 `SUB_AGENT_SHARED_WRITE_FORBIDDEN`。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-git-and-shell-write-commands.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Git And Shell Write Commands\n\n## Summary\n- 结论：只读 sub-agent 不得执行 Git ref/remote 写操作或命令式文件写入。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 覆盖底层写入命令\n  checker 必须识别 Git push、branch、tag、update-ref 与 sed、tee、重定向。\n\n## Non-goals\n- 不开放 sub-agent 的任何写入能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:push [gpt-5.6-luna, xhigh] => git push (skill: $normal-flow)\n  - sub-agent:branch [gpt-5.6-luna, xhigh] => git branch feature (skill: $normal-flow)\n  - sub-agent:tag [gpt-5.6-luna, xhigh] => git tag release (skill: $normal-flow)\n  - sub-agent:ref [gpt-5.6-luna, xhigh] => git update-ref refs/heads/feature HEAD (skill: $normal-flow)\n  - sub-agent:sed [gpt-5.6-luna, xhigh] => sed -i changes current file (skill: $normal-flow)\n  - sub-agent:tee [gpt-5.6-luna, xhigh] => tee current-file (skill: $normal-flow)\n  - sub-agent:redirection [gpt-5.6-luna, xhigh] => printf content > current-file (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有 Git 和 shell 写入命令。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-git-config-worktree.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Git Config And Worktree Writes\n\n## Summary\n- 结论：只读 sub-agent 不得修改 Git 配置或管理 worktree。\n\n## Flow Decision\n- flow_type: normal\n- flow_source: user_explicit\n- lock_status: locked\n- 开发基线：origin/main\n\n## Goals\n- 目标 1: 覆盖 Git 配置和 worktree 写入\n  checker 必须拒绝会改变配置或 worktree 生命周期的命令。\n\n## Non-goals\n- 不开放 sub-agent 的 Git 写入能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:config [gpt-5.6-luna, xhigh] => git config --local user.name Agent (skill: $normal-flow)\n  - sub-agent:add-worktree [gpt-5.6-luna, xhigh] => git worktree add ../other-worktree branch (skill: $normal-flow)\n  - sub-agent:remove-worktree [gpt-5.6-luna, xhigh] => git worktree remove ../other-worktree (skill: $normal-flow)\n  - sub-agent:config-chain [gpt-5.6-luna, xhigh] => git config --get user.name && git config --global user.name Agent (skill: $normal-flow)\n  - sub-agent:apply-chain [gpt-5.6-luna, xhigh] => git apply --check patch.diff || git apply patch.diff (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有 Git 配置和 worktree 写入命令，包括链式命令。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-git-global-options.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Git Global Options\n\n## Summary\n- 结论：只读 sub-agent 不得执行裸 commit 或带 Git 全局选项的写命令。\n\n## Flow Decision\n- flow_type: normal\n- flow_source: user_explicit\n- lock_status: locked\n- 开发基线：origin/main\n\n## Goals\n- 目标 1: 覆盖 Git 命令变体\n  checker 必须拒绝裸 commit、-C 路径和 --git-dir 形式的写命令。\n\n## Non-goals\n- 不开放 sub-agent 的 Git 写入能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:bare [gpt-5.6-luna, xhigh] => git commit (skill: $normal-flow)\n  - sub-agent:global-add [gpt-5.6-luna, xhigh] => git -C . add current-file (skill: $normal-flow)\n  - sub-agent:global-push [gpt-5.6-luna, xhigh] => git --git-dir=/tmp/repo push (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有命令变体。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-git-readonly-mixed-commands.jsonl

@@ -0,0 +1 @@
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Mixed Readonly Git Command\n\n## Summary\n- 结论：只读 Git 查询与写命令混在同一 sub-agent task 中必须被拒绝。\n\n## Flow Decision\n- flow_type: normal\n- flow_source: user_explicit\n- lock_status: locked\n- 开发基线：origin/main\n\n## Goals\n- 目标 1: 拒绝混合命令\n  checker 必须识别只读查询后跟随的写入命令。\n\n## Non-goals\n- 不开放 sub-agent 写入共享工作区。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:config-push [gpt-5.6-luna, xhigh] => inspect git config --get user.name before git push origin main (skill: $normal-flow)\n\n## Acceptance & Validation\n- 混合只读与写命令必须失败。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-git-write-commands.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Git Write Commands\n\n## Summary\n- 结论：只读 sub-agent 不得执行会修改 index、工作树或 refs 的 Git 命令。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 覆盖 Git 写入命令\n  checker 必须拒绝 add、stage、checkout、restore 和 apply。\n\n## Non-goals\n- 不开放 sub-agent 的任何 Git 写入能力。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:add [gpt-5.6-luna, xhigh] => git add files in the repository (skill: $normal-flow)\n  - sub-agent:stage [gpt-5.6-luna, xhigh] => git stage changes in the worktree (skill: $normal-flow)\n  - sub-agent:checkout [gpt-5.6-luna, xhigh] => git checkout the current worktree file (skill: $normal-flow)\n  - sub-agent:restore [gpt-5.6-luna, xhigh] => git restore files in the repository (skill: $normal-flow)\n  - sub-agent:apply [gpt-5.6-luna, xhigh] => git apply the patch (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有 Git 写入命令。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-implementation-multiword.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划多词实现边界"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Multiword Implementation Assignment\n\n## Summary\n- 结论：sub-agent 不能承接多词实现任务。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 拦截多词实现任务\n  checker 必须拒绝 implement the bug fix、implement the code changes 与 implement this feature。\n\n## Non-goals\n- 不开放 sub-agent 的任何实现权限。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:bug-fix [gpt-5.6-luna, xhigh] => implement the bug fix (skill: $normal-flow)\n  - sub-agent:code-changes [gpt-5.6-luna, xhigh] => implement the code changes (skill: $normal-flow)\n  - sub-agent:feature [gpt-5.6-luna, xhigh] => implement this feature (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须报告 `SUB_AGENT_SHARED_WRITE_FORBIDDEN`。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-implementing-write.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按 fast-track 规划英文只读边界。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Implementing Write\n\n## Summary\n- 结论：英文动词变形也必须遵守 sub-agent 只读边界。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n- 收口条件：`merge-ready / Step 5C Ready`\n\n## Goals\n- 目标 1: 拦截 implementing 写入声明\n  checker 必须拒绝把当前 worktree 作为实现目标的 sub-agent。\n\n## Non-goals\n- 不开放 sub-agent 的写入权限。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:impl [gpt-5.6-luna, xhigh] => implementing the source code in the current worktree (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 必须报告 `SUB_AGENT_SHARED_WRITE_FORBIDDEN`。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-local-write.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Task Local Write\n\n## Summary\n- 结论：sub-agent 不得写入 task-local worktree 或其他文件系统目标。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- 目标 1: 验证完整只读边界\n  checker 必须拒绝 task-local worktree 和 another filesystem 的写入声明。\n\n## Non-goals\n- 不开放 sub-agent 的任何写入目标。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:review [gpt-5.6-luna, xhigh] => write to a task-local worktree (skill: $normal-flow)\n  - sub-agent:test [gpt-5.6-luna, xhigh] => modify another filesystem location (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝所有非共享工作区写入。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-shared-write-english.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid English Shared Write\n\n## Summary\n- 结论：English shared-workspace writes must remain forbidden.\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证英文只读边界\n  The checker must reject writes to the shared workspace and modifications to the current worktree.\n\n## Non-goals\n- 不允许并行代码写入。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:review [gpt-5.6-luna, xhigh] => write to the shared workspace and commit changes (skill: $normal-flow)\n  - sub-agent:impl [gpt-5.6-luna, xhigh] => modify the current worktree (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker must reject both shared-write declarations.\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-subagent-shared-write.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Shared Write\n\n## Summary\n- 结论：sub-agent 声明写入共享工作区。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- 目标 1: 验证只读边界\n  sub-agent 只能回传检查结果。\n\n## Non-goals\n- 不允许并行代码写入。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:review [gpt-5.6-luna, xhigh] => 写入共享工作区并提交修复 (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝共享工作区写入。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-task-agent-alias.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"走 fast-track"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Task Agent Alias\n\n## Summary\n- 结论：旧 task-agent 别名不得出现在新计划。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证 executor 命名\n  只允许 task:<name> 表示独立 Codex task。\n\n## Non-goals\n- 不保留新旧 executor 双重输出。\n\n## Task Orchestration\n- wave: 1\n  - task-agent:ticket-one [gpt-5.6-luna, xhigh] => 创建独立 task (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝 task-agent 别名。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-task-executor-noninitiative.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"走 fast-track"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Direct Task\n\n## Summary\n- 结论：direct flow 错误地创建独立 task。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证 flow 边界\n  fast-track 不得创建 Initiative task Worktree。\n\n## Non-goals\n- 不引入跨 Ticket 并行。\n\n## Task Orchestration\n- wave: 1\n  - task:ticket-one [gpt-5.6-luna, xhigh] => 创建独立 Ticket task (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝 direct flow 的 task executor。\n\n## Assumptions/Defaults\n- 用于负例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-ticket-replans-via-executor.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 走 fast-track。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Ticket Executor Replan\n\n## Summary\n- 结论：Ticket executor 错误地重开了计划。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 阻止 Ticket 再规划\n  Ticket 必须继承批准计划。\n\n## Non-goals\n- 不创建新的计划会话。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:ticket [gpt-5.6-luna, xhigh] => 实现已批准的范围 (skill: $dev-plan + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须从 executor 身份识别 Ticket 重规划。\n\n## Assumptions/Defaults\n- 该 Ticket 已经拥有批准计划：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/negative-ticket-replans.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 走 fast-track。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Invalid Ticket Replan\n\n## Summary\n- 结论：Ticket 会错误地再次规划。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 实现 Ticket\n  Ticket 在实现前不应再次创建计划。\n\n## Non-goals\n- 不创建 initiative integration branch。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:ticket [gpt-5.6-luna, xhigh] => Ticket 实现前调用 $dev-plan 重新规划 (skill: $dev-plan + $fast-flow)\n\n## Acceptance & Validation\n- Ticket 完成实现。\n\n## Assumptions/Defaults\n- 已存在批准计划：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/non-baseline-based-on-docs-text.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 基于 `docs/specs/7uqp8-mainline-baseline-topic-branch-freshness/SPEC.md` 规划这个任务，并给出 `flow_type: fast-track` 计划。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Based-on Spec Text Is Not A Git Baseline\n\n## Summary\n- 结论：普通“基于某个 spec 文档制定计划”不应被当成主人显式指定 Git baseline。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证 generic based-on 文案不会触发 owner baseline mismatch\n  当前请求只是说明 planning 输入来自某个 spec 文档，不代表主人显式指定了 Git branch/ref/commit baseline。\n\n## Non-goals\n- 不覆盖真正的显式 branch/commit baseline 场景。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于当前 spec 输入的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 不得因为 `基于 \\`docs/specs/.../SPEC.md\\`` 这类 wording 误报 `OWNER_BASELINE_MISMATCH`。\n\n## Assumptions/Defaults\n- 主人没有显式给出 Git baseline，只是给了 spec 输入文档：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/non-goals-blank-bullet.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "按普通流程计划"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Non-goals Blank Bullet\n\n## Summary\n- 结论：用于验证空 non-goals bullet 不能绕过 goals-first 合同。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary and success signal before execution arrangement.\n\n## Non-goals\n-\n\n## Task Orchestration\n- wave: 1\n  - main-agent => run local validation only (skill: $normal-flow)\n\n## Acceptance & Validation\n- Replay fixture must fail before execution starts.\n\n## Assumptions/Defaults\n- 该样例专门用于验证空 non-goals bullet：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/non-goals-none-placeholder.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "按普通流程计划"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Non-goals None Placeholder\n\n## Summary\n- 结论：用于验证 non-goals 里的 `None` 不能充当具体排除项。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary and success signal before execution arrangement.\n\n## Non-goals\n- None\n\n## Task Orchestration\n- wave: 1\n  - main-agent => run local validation only (skill: $normal-flow)\n\n## Acceptance & Validation\n- Replay fixture must fail before execution starts.\n\n## Assumptions/Defaults\n- 该样例专门用于验证 non-goals placeholder：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/non-goals-placeholder.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"按普通流程计划"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Non-goals Placeholder\n\n## Summary\n- 结论：用于验证 Non-goals 占位 bullet 会被拦截。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- Goal 1: lock a concrete scope\n  Keep the plan focused on one concrete implementation target.\n\n## Non-goals\n- TBD\n\n## Task Orchestration\n- wave: 1\n  - main-agent => implement the scoped change (skill: $normal-flow)\n\n## Acceptance & Validation\n- Local validation proves the scoped change behaves as expected.\n\n## Assumptions/Defaults\n- Scope boundary is already understood: 已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/persisted-plan-artifact-in-orchestration.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "按 normal 流程计划"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Persisted Plan Artifact in Orchestration\n\n## Summary\n- 结论：用于验证 Task Orchestration 禁止落盘计划载体。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- Goal 1: preserve non-persistent planning boundaries\n  The task arrangement must not create PLAN.md, PLANS.md, or any progress ledger artifact.\n\n## Non-goals\n- Do not permit persisted plan files in orchestration.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => write PLAN.md progress ledger before implementing the checker change (skill: $normal-flow)\n\n## Acceptance & Validation\n- The checker must reject persisted plan artifact work inside Task Orchestration.\n\n## Assumptions/Defaults\n- This fixture intentionally exercises the persisted-artifact ban: 已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/placeholder-acceptance.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "按 normal 流程计划"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Placeholder Acceptance\n\n## Summary\n- 结论：用于验证占位式 Acceptance & Validation 会被拦截。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- Goal 1: keep acceptance actionable\n  Plans must provide a concrete validation line instead of deferring acceptance details to later.\n\n## Non-goals\n- Do not allow placeholder-only acceptance text.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => implement the demo workflow change (skill: $normal-flow)\n\n## Acceptance & Validation\n- TBD later.\n\n## Assumptions/Defaults\n- This fixture intentionally exercises placeholder acceptance wording: 已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/placeholder-development-baseline.jsonl

@@ -0,0 +1 @@
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Placeholder Development Baseline\n\n## Summary\n- 结论：示例故意把开发基线写成占位词。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`TBD`\n\n## Goals\n- 目标 1: 演示 checker 不能把占位词当成 concrete baseline\n  当前计划故意把 `TBD` 填进 `开发基线`，用于验证 checker 会拒绝这类 placeholder 值。\n\n## Non-goals\n- 不覆盖真正的 branch/ref、commit 或抽象主干词场景。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须拒绝把 `TBD` 这类 placeholder 当成 `Flow Decision` 的有效 `开发基线`。\n\n## Assumptions/Defaults\n- 这是一个专门用于占位 baseline 校验的反例：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/plan-md-handoff-job-future-sessions-artifact.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "按 normal 流程计划"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# PLAN.md Handoff Job Future Sessions Artifact\n\n## Summary\n- 结论：用于验证带 future-session planning context 的 `PLAN.md` job 仍会被拦截。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n\n## Goals\n- Goal 1: preserve non-persistent planning boundaries\n  The orchestration must not write a persisted `PLAN.md` handoff artifact for future sessions.\n\n## Non-goals\n- Do not allow planning `PLAN.md` handoff jobs inside Task Orchestration.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => write PLAN.md handoff job for future sessions (skill: $normal-flow)\n\n## Acceptance & Validation\n- The checker rejects `PLAN.md` handoff jobs when they target future sessions.\n\n## Assumptions/Defaults\n- This fixture intentionally exercises planning-context `PLAN.md` job detection: 已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-agent-profiles.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"走 fast-track"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Agent Profile Contract\n\n## Summary\n- 结论：验证 direct flow 的 sub-agent profile。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证 profile\n  让只读子代理使用可解析的模型和 effort。\n\n## Non-goals\n- 不创建独立 Ticket task。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:review [gpt-5.6-luna, xhigh] => 只读分析当前变更 (skill: $fast-flow)\n- wave: 2\n  - sub-agent:hard [gpt-5.6-luna, max] => 只读处理最复杂的验证 (skill: $fast-flow)\n- wave: 3\n  - main-agent => 汇总结果并准备 PR (skill: $fast-flow)\n\n## Acceptance & Validation\n- checker 接受带 profile 的 sub-agent。\n\n## Assumptions/Defaults\n- profile 合法：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-default-fast-track-no-flow.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请为这个需求直接产出可执行计划。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Default Fast-track Auto-lock\n\n## Summary\n- 结论：当前请求没有显式 flow 信号，按默认 fast-track + merge-ready 规划。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 直接落默认快车道\n  在没有显式 flow 指令且不存在歧义 blocker 时，直接进入 fast-track 规划，避免为 happy path 再弹一次 flow 选择。\n\n## Non-goals\n- 不把 generic `继续` 或互斥 flow 信号静默当成默认快车道。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划实现、验证、PR 收敛与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- 默认路径直接生成 `fast-track` 计划，并把 stop condition 锁到 `merge-ready / Step 5C Ready`。\n\n## Assumptions/Defaults\n- 当前请求未显式指定其他 flow：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-default-initiative-from-planned-scope.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 为账号管理和计费能力制定开发计划。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Account And Billing Initiative\n\n## Summary\n- 结论：按计划范围的独立交付边界选择 Initiative。\n\n## Flow Decision\n- `flow_type`: `initiative`\n- `flow_source`: `default_initiative_complexity`\n- `lock_status`: `locked`\n- `integration_branch`: `prd/account-billing`\n- `final_base`: `main`\n- `schema_version`: `5`\n- `initiative_contract_version`: `5`\n- `promotion_policy`: `checkpointed`\n- `integration_ci_surface`: `branch-push`\n- `integration_ci_check`: `quality`\n- `plan_kind`: `delivery`\n- `spec_disposition`: `reuse`\n- `spec_path`: `docs/specs/account-billing/SPEC.md`\n- `adr_contracts`: `none`\n- `tracking_issue_url`: `https://github.com/acme/example/issues/100`\n- `bootstrap_contract_digest`: `aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa`\n- `ticket_topology_source`: `to-tickets`\n- `ticket_topology_digest`: `bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb`\n- `ticket_source_sha`: `1111111111111111111111111111111111111111`\n- `native_task_budget`: `2`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 建立账号管理 Ticket\n  账号管理与计费形成至少两个独立验收、独立合并的 Ticket。\n- 目标 2: 建立计费 Ticket\n  两个 Ticket 通过隔离 task Worktree 和 Wave 汇入 integration branch。\n\n## Non-goals\n- 不自动合并 aggregate PR。\n\n## Task Orchestration\n- wave: 1\n  - task:issue-101 [gpt-5.6-terra, xhigh] => 实现并验证账号管理 Ticket (skill: $initiative-flow)\n- wave: 2\n  - task:issue-102 [gpt-5.6-terra, xhigh] => 实现并验证计费 Ticket (skill: $initiative-flow)\n\n## Acceptance & Validation\n- task:issue-101 => acceptance_scope=technical-only; evidence=integration-run@account-task; owner_gate=none; completion_gate=integration\n- task:issue-102 => acceptance_scope=technical-only; evidence=integration-run@billing-task; owner_gate=none; completion_gate=integration\n\n## Assumptions/Defaults\n- 计划范围证明存在两个独立交付 Ticket：已确定\n- `$to-tickets` topology 已获批准并已发布：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-default-mainline-concrete-baseline.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我可执行计划。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Default Mainline Concrete Baseline\n\n## Summary\n- 结论：未显式指定 base 时，默认按目标仓库主干规划，但最终仍要输出具体 ref。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 验证默认主干策略会落成具体 ref\n  当前请求没有显式 branch/commit，也没有其他强 Git 事实，因此最终 `开发基线` 应写成 concrete ref，而不是抽象主干词。\n\n## Non-goals\n- 不覆盖显式 owner baseline 的保留路径。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划实现、验证与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受“默认主干策略 + concrete ref 输出”的计划结果。\n\n## Assumptions/Defaults\n- 当前请求没有显式指定其他 base：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-default-mainline-token-ref.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `feature/default-mainline-baseline-cleanup`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Default Mainline Token Ref\n\n## Summary\n- 结论：合法分支名即使包含 `default` 或 `mainline` 词根，也仍然是 concrete baseline。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`feature/default-mainline-baseline-cleanup`\n\n## Goals\n- 目标 1: 接受包含默认主干词根的真实 ref\n  当前计划故意使用 `feature/default-mainline-baseline-cleanup` 作为最终 ref，用于验证 checker 不会把真实分支名误判为抽象主干策略词。\n\n## Non-goals\n- 不把词根相似的真实 ref 视为占位词或抽象策略短语。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于真实 feature ref 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受带有 `default/mainline` 词根的合法 git ref。\n\n## Assumptions/Defaults\n- 主人已显式给出带默认主干词根的合法 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-direct-dev-plan-issue-parser.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 走 fast-track。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Direct Issue Parser Maintenance\n\n## Summary\n- 结论：这是 direct 单 PR 维护，不是继承计划的 Ticket 执行。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 修正 direct 计划中的 Issue 解析\n  `main-agent` 维护 `$dev-plan` 的 Issue/Ticket 解析能力，同时保留单 PR 交付语义。\n\n## Non-goals\n- 不把该 direct 维护任务转换为 initiative Ticket 执行。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => update the $dev-plan Issue parser for direct-flow recovery (skill: $dev-plan + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受包含 Issue/Ticket 文本的 direct `$dev-plan` 维护任务。\n\n## Assumptions/Defaults\n- 当前任务不是继承批准计划的 Ticket：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-direct-mentions-initiative-non-goal.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 走 fast-track，规划一个单 PR 交付。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Direct Delivery With Initiative Boundary\n\n## Summary\n- 结论：这是 direct 单 PR 交付，停在 merge-ready。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 完成单 PR 交付\n  在一个分支中实现、验证并收敛该范围。\n\n## Non-goals\n- 不调用 `$initiative-flow`，也不创建 Ticket 或 aggregate PR。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 实现并验证 direct 交付 (skill: $fast-flow)\n- wave: 2\n  - main-agent => 创建并收敛 direct PR (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- 该 PR 的最新 head 完成验证后停在 merge-ready。\n\n## Assumptions/Defaults\n- 这是单交付面需求：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-directional-product-wording-no-baseline.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 基于 new plan file upload API 规划这个任务。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Directional Product Wording Does Not Create Baseline\n\n## Summary\n- 结论：产品词组里的 directional wording 不应被误抽成主人显式开发基线。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 证明产品短语 `new plan file upload API` 不会被误识别为 baseline\n  这里的 `基于 new plan file upload API` 描述的是任务范围，而不是 git branch/ref/commit，因此计划应回到目标仓库主干的具体 ref。\n\n## Non-goals\n- 不把普通产品命名误判成 owner-explicit baseline。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划 plan file upload API 的实现、验证与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 directional 产品词组不会触发 owner baseline 提取。\n\n## Assumptions/Defaults\n- 主人没有显式指定 branch 或 commit：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-equivalent-remote-ref-baseline.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `origin/main`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Equivalent Remote Ref Baseline\n\n## Summary\n- 结论：等价的 remote ref 完整写法不应被误报成 owner baseline mismatch。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`refs/remotes/origin/main`\n\n## Goals\n- 目标 1: 接受与 owner baseline 等价的 remote ref 完整写法\n  当前计划使用 `refs/remotes/origin/main` 表达与主人显式 `origin/main` 等价的基线，不应被当成错误覆盖。\n\n## Non-goals\n- 不把真正不同的 branch/ref 当成等价值。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 remote mainline 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 `origin/main` 与 `refs/remotes/origin/main` 的 owner baseline 等价比较。\n\n## Assumptions/Defaults\n- 主人已显式给出 remote branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-after-flow-answer.jsonl

@@ -0,0 +1,4 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，开发基线使用 `upstream/release-2026q3`。"}]}}
+{"type":"response_item","payload":{"type":"function_call","name":"request_user_input","call_id":"call_flow_type_lock","arguments":"{\"questions\":[{\"header\":\"流程类型\",\"id\":\"flow_type\",\"question\":\"当前 flow 信号存在歧义，请明确本次流程类型\",\"options\":[{\"label\":\"快车道 (Recommended)\",\"description\":\"推进到 PR merge-ready / Step 5C Ready\"},{\"label\":\"仅实现\",\"description\":\"只产出实现规格\"},{\"label\":\"普通流程\",\"description\":\"验证后本地提交并停下\"},{\"label\":\"本地合主干\",\"description\":\"规划到本地 merge-ready 与本地合并收尾。\"}]}]}"}}
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"normal"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline After Flow Answer\n\n## Summary\n- 结论：即使流程类型是后续单独回答的，主人先前显式给出的 branch baseline 也必须被原样保留。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 在 flow 回答后继续保留显式 branch baseline\n  当前计划应在 `normal` 流程下继续围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把后续 flow 回答误当成新的 baseline 清空点。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划普通流程下的实现、验证与本地提交收口 (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须接受 flow 回答后的显式 branch baseline 被原样保留。\n\n## Assumptions/Defaults\n- 主人已先给出 branch baseline，流程类型在后续单独回答：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-after-prose-flow-answer.jsonl

@@ -0,0 +1,4 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，开发基线使用 `upstream/release-2026q3`。"}]}}
+{"type":"response_item","payload":{"type":"function_call","name":"request_user_input","call_id":"call_flow_type","arguments":"{\"questions\":[{\"header\":\"流程类型\",\"id\":\"flow_type\",\"question\":\"当前 flow 信号存在歧义，请明确本次流程类型\",\"options\":[{\"label\":\"快车道 (Recommended)\",\"description\":\"推进到 PR merge-ready / Step 5C Ready\"},{\"label\":\"仅实现\",\"description\":\"只产出实现规格\"},{\"label\":\"普通流程\",\"description\":\"验证后本地提交并停下\"},{\"label\":\"本地合主干\",\"description\":\"规划到本地 merge-ready 与本地合并收尾。\"}]}]}"}}
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请按 normal 继续出计划。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline After Prose Flow Answer\n\n## Summary\n- 结论：即使流程类型是通过自然语言补答的，主人先前显式给出的 branch baseline 也必须被原样保留。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 在 prose flow answer 后继续保留显式 branch baseline\n  当前计划应在 `normal` 流程下继续围绕 `upstream/release-2026q3` 编排，而不是因为 flow 问答后的 session reset 把 baseline 清空。\n\n## Non-goals\n- 不把自然语言的 flow 回答误判成新的 planning window 清空点。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划普通流程下的实现、验证与本地提交收口 (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须接受 prose flow answer 后的显式 branch baseline 被原样保留。\n\n## Assumptions/Defaults\n- 主人已先给出 branch baseline，流程类型在后续自然语言中补答：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-after-standalone-flow-restatement.jsonl

@@ -0,0 +1,4 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: normal` 计划，开发基线使用 `upstream/release-2026q3`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline After Standalone Flow Restatement\n\n## Summary\n- 结论：即使主人在已锁定计划后单独重述 flow，先前显式给出的 branch baseline 也必须被原样保留。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 在 standalone flow restatement 后继续保留显式 branch baseline\n  当前计划应在 `normal` 流程下继续围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把 standalone flow restatement 误当成新的 baseline 清空点。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划普通流程下的实现与本地提交收口 (skill: $normal-flow)\n\n## Acceptance & Validation\n- 第一轮计划必须原样保留显式 branch baseline。\n\n## Assumptions/Defaults\n- 第一轮请求已显式给出 branch baseline：已确定\n</proposed_plan>"}]}}
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"normal"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline After Standalone Flow Restatement Revision\n\n## Summary\n- 结论：修订计划时即使主人只单独重述 flow，显式 branch baseline 也必须继续保留。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 修订计划时继续保留显式 branch baseline\n  后续单独重述 `normal` flow 不应改变当前围绕 `upstream/release-2026q3` 的编排约束。\n\n## Non-goals\n- 不把 standalone flow restatement 误当成新的 baseline 清空点。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划普通流程下的实现与本地提交收口 (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须接受 standalone flow restatement 后显式 branch baseline 被原样保留。\n\n## Assumptions/Defaults\n- standalone flow restatement 不应清空第一轮显式 baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-base-ref-equals.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，BASE_REF=upstream/release-2026q3。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline BASE_REF Equals\n\n## Summary\n- 结论：主人用 `BASE_REF=...` 显式给出 branch baseline 时，计划也必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留 `BASE_REF=...` 形式的显式 branch baseline\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把 `BASE_REF=...` 忽略成默认主干解析。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 `BASE_REF=...` branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已用 `BASE_REF=...` 显式给出 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-directional-baseline.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，基于 `upstream/release-2026q3` 开发。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Directional Wording\n\n## Summary\n- 结论：主人即使用自然语句表达“基于某个 branch 开发”，计划也必须原样保留该 baseline。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留 directional wording 形式的显式 branch baseline\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把 natural-language baseline 忽略成默认主干解析。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 directional wording branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已用 directional wording 显式给出 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-doc-like-ref.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `feature/README.md`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Doc-like Ref\n\n## Summary\n- 结论：当主人显式给出的 baseline 是合法 branch ref `feature/README.md` 时，计划必须原样保留这个具体开发基线。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`feature/README.md`\n\n## Goals\n- 目标 1: 接受长得像文档路径的合法 branch ref\n  这个计划用于验证 checker 会优先接受合法 git ref，而不是因为 ref 末尾像 `README.md` 就把它误判成仓库文件路径。\n\n## Non-goals\n- 不放开真正的文件系统路径，例如 `./docs/specs/foo/SPEC.md`。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 `feature/README.md` 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 `feature/README.md` 作为具体开发基线。\n\n## Assumptions/Defaults\n- 主人已显式给出长得像文档路径但合法的 branch ref baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-feature-head.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 基于 feature/HEAD 规划这个任务"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Feature HEAD\n\n## Summary\n- 结论：当主人显式给出的 baseline 是合法 branch ref `feature/HEAD` 时，计划必须原样保留这个具体开发基线。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`feature/HEAD`\n\n## Goals\n- 目标 1: 接受以 `HEAD` 结尾的合法 branch ref\n  这个计划用于验证 checker 不会把合法 branch 名误判成 context-dependent baseline。\n\n## Non-goals\n- 不放开真正的 symbolic HEAD alias，例如 `origin/HEAD`。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 `feature/HEAD` 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 `feature/HEAD` 作为具体开发基线。\n\n## Assumptions/Defaults\n- 主人已显式给出合法的 branch ref baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-legal-trailing-punct-comma-clause.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，开发基线使用 release/v1.0!，并继续。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Legal Trailing Punctuation With Comma Clause\n\n## Summary\n- 结论：plain text baseline 即使本体以合法标点字符结尾，后面再跟逗号续句时也必须保留真实 branch baseline。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`release/v1.0!`\n\n## Goals\n- 目标 1: 保留逗号续句前的合法尾部标点 branch baseline\n  当前计划应直接围绕 `release/v1.0!` 编排，而不是把 `，并继续` 一起吞进 baseline，或静默裁掉合法尾部字符。\n\n## Non-goals\n- 不把合法尾部标点误当成自然语言句尾，也不把后续逗号续句当成 baseline 内容。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受带合法尾部标点且后接逗号续句的 plain text branch baseline 被正确截断并原样保留。\n\n## Assumptions/Defaults\n- 主人已在逗号续句前显式给出合法 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-legal-trailing-punct.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `release/v1.0!`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Legal Trailing Punctuation\n\n## Summary\n- 结论：主人显式给出尾部带合法标点字符的 branch baseline 时，计划必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`release/v1.0!`\n\n## Goals\n- 目标 1: 保留主人显式指定的合法尾部标点 branch baseline\n  当前计划应直接围绕 `release/v1.0!` 编排，而不是静默删除尾部字符。\n\n## Non-goals\n- 不把合法尾部标点误当成自然语言句尾并裁掉。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受带合法尾部标点字符的 branch baseline 被原样保留。\n\n## Assumptions/Defaults\n- 主人已显式给出合法尾部标点 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-legal-trailing-semicolon-plain.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 release/v1;"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Plain Legal Trailing Semicolon\n\n## Summary\n- 结论：主人显式给出尾部带合法分号字符的 plain branch baseline 时，计划必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`release/v1;`\n\n## Goals\n- 目标 1: 保留主人显式指定的 plain 合法尾部分号 branch baseline\n  当前计划应直接围绕 `release/v1;` 编排，而不是静默删除尾部合法字符。\n\n## Non-goals\n- 不把 plain baseline 末尾的合法分号误当成句尾标点裁掉。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受尾部带合法分号字符的 plain branch baseline 被原样保留。\n\n## Assumptions/Defaults\n- 主人已显式给出 plain 合法尾部分号 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-placeholder-token-ref.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `todo`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Placeholder Token Ref\n\n## Summary\n- 结论：当主人显式给出的 baseline 本身就是合法 branch ref 时，哪怕名字与 placeholder 词根重合，也必须被当作 concrete ref 保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`todo`\n\n## Goals\n- 目标 1: 接受与 placeholder 词根重合的合法 branch ref\n  当前计划故意使用 `todo` 作为最终 ref，用于验证 checker 会优先接受 concrete git ref，而不是先被 placeholder 词表误伤。\n\n## Non-goals\n- 不放开真正的 placeholder 值，例如 `TBD`。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 `todo` branch 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受与 placeholder 词根重合的合法 branch baseline。\n\n## Assumptions/Defaults\n- 主人已显式给出与 placeholder 词根重合的合法 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-plain-chinese.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，开发基线使用 upstream/release-2026q3。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Plain Chinese\n\n## Summary\n- 结论：主人用中文 plain text 指定 branch baseline 时，计划也必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留中文 plain text 形式的显式 branch baseline\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把中文 plain text baseline 忽略成默认主干解析。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受中文 plain text branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已用中文 plain text 显式给出 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-replace-clause.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线从 `origin/main` 改为 `upstream/release-2026q3`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Replace Clause\n\n## Summary\n- 结论：当主人在同一句里把开发基线从旧值改为新值时，计划必须以新的显式 baseline 为准。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留 replace clause 中的新 branch baseline\n  当前计划必须围绕 `upstream/release-2026q3` 编排，而不是错误沿用句子里被替换掉的 `origin/main`。\n\n## Non-goals\n- 不把被替换的旧 baseline 当成最终开发基线。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于新 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受 `开发基线从 A 改为 B` 这类句型，并把最终显式 baseline 识别为 `B`。\n\n## Assumptions/Defaults\n- 主人已在同一句里把开发基线改为新的 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-short-ref.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 qa"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Short Ref\n\n## Summary\n- 结论：主人显式给出短 branch baseline 时，计划必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`qa`\n\n## Goals\n- 目标 1: 保留主人显式指定的短 branch baseline\n  当前计划应直接围绕 `qa` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把合法短 ref 误当成缺失或无效 baseline。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于短 branch ref 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受短 ref branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已显式给出短 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-special-chars.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `release/v1.2+hotfix`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline With Special Characters\n\n## Summary\n- 结论：主人显式给出包含合法 git ref 特殊字符的 branch baseline 时，计划必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`release/v1.2+hotfix`\n\n## Goals\n- 目标 1: 保留主人显式指定的合法 branch baseline\n  当前计划应直接围绕 `release/v1.2+hotfix` 编排，而不是因为 checker 的字符预筛选被误拒绝。\n\n## Non-goals\n- 不把合法 git ref 特殊字符误判成非法 baseline。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release hotfix 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受包含 `+` 的合法 branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已显式给出合法的 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-topic-prefix.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `topic/foo`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Topic Prefix\n\n## Summary\n- 结论：即使 branch ref 自身包含 `to` 词根，只要它是合法 git ref，显式 baseline 也必须被原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`topic/foo`\n\n## Goals\n- 目标 1: 保留包含 `to` 词根的显式 branch baseline\n  当前计划应直接围绕 `topic/foo` 编排，而不是因为 replace-clause 词根误匹配而把 owner baseline 截断。\n\n## Non-goals\n- 不把 branch ref 内部的 `to` 词根误当成 baseline 替换指令。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 `topic/foo` 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受包含 `to` 词根的合法 branch baseline 被原样保留。\n\n## Assumptions/Defaults\n- 主人已显式给出包含 `to` 词根的 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-trailing-comma-clause.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，开发基线使用 upstream/release-2026q3, 并继续。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Trailing Comma Clause\n\n## Summary\n- 结论：中文 plain text baseline 后面即使跟着逗号分句，计划也必须保留真实 branch baseline。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留逗号分句前的显式 branch baseline\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是把后续分句一起吞进 baseline。\n\n## Non-goals\n- 不把 `并继续` 这类后续短语当成 baseline 内容。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受逗号分句后的 plain text branch baseline 被正确截断并原样保留。\n\n## Assumptions/Defaults\n- 主人已在逗号分句前显式给出 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline-yong-plain-chinese.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 规划这项工作，开发基线用 upstream/release-2026q3。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline Yong Plain Chinese\n\n## Summary\n- 结论：主人用中文 `开发基线用 ...` 指定 branch baseline 时，计划也必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留中文 `开发基线用 ...` 形式的显式 branch baseline\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不把中文 `开发基线用 ...` baseline 忽略成默认主干解析。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受中文 `开发基线用 ...` branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已用中文 `开发基线用 ...` 显式给出 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-branch-baseline.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `upstream/release-2026q3`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Branch Baseline\n\n## Summary\n- 结论：主人显式给出 branch baseline 时，计划必须原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留主人显式指定的 branch baseline\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是回退到默认主干 ref。\n\n## Non-goals\n- 不回退成目标仓库主干的默认解析。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受显式 branch baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已显式给出 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-commit-baseline.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请用 $dev-plan 直接给我 `flow_type: fast-track` 计划，开发基线使用 `140953065ecfe450b870053f7e4e3d02010cf45a`。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit Commit Baseline\n\n## Summary\n- 结论：主人显式给出 commit baseline 时，计划必须原样保留该 SHA。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`140953065ecfe450b870053f7e4e3d02010cf45a`\n\n## Goals\n- 目标 1: 保留主人显式指定的 commit baseline\n  当前计划应围绕该确定提交组织后续执行，而不是改写成 branch 名称。\n\n## Non-goals\n- 不把 commit baseline 自动替换成主干 branch。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于固定 commit 的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受显式 commit baseline 被原样保留的计划输出。\n\n## Assumptions/Defaults\n- 主人已显式给出 commit baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-explicit-english-branch-baseline-is.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"Please use $dev-plan with `flow_type: fast-track` and keep the base branch is upstream/release-2026q3."}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Explicit English Baseline With Is\n\n## Summary\n- 结论：英文 `base branch is ...` 这类显式 owner baseline 也必须被原样保留。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`upstream/release-2026q3`\n\n## Goals\n- 目标 1: 保留英文显式 baseline 文案\n  当前计划应直接围绕 `upstream/release-2026q3` 编排，而不是把英文 `base branch is ...` 忽略掉。\n\n## Non-goals\n- 不回退成默认主干 ref。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划基于 release 分支的实现与 merge-ready 收口 (skill: $review-agent + $fast-flow)\n\n## Acceptance & Validation\n- checker 必须接受英文 `base branch is ...` 形式的显式 baseline 被原样保留。\n\n## Assumptions/Defaults\n- 主人已显式给出英文 branch baseline：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-fast-track.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"走快车道（fast-track）"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# 快车道示例计划\n\n## Summary\n- 结论：快车道编排包含分支、PR、review 收敛。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- `review_tier`: `Tier 3`\n- `review_rationale`: `current-head proof is required for the changed workflow contract`\n- `runtime_review_strategy`: `launch fresh formal lanes only after Review Readiness and current-candidate validation are complete`\n- `review_profile_recommendation`: `recommend gpt-6-luna + high effort for four to six formal lanes`\n- 开发基线：`origin/main`\n- 收口条件：`merge-ready / Step 5C Ready`\n\n## Goals\n- 目标 1: 锁定示例范围合同\n  在执行编排前明确结果边界、关键实现细节与成功信号。\n\n## Non-goals\n- 不扩展示例合同之外的范围。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 完成 readiness、验证和 PR 收口 (skill: $fast-flow)\n\n## Acceptance & Validation\n- A1 [contract/direct] runtime review | evidence=`test:review` | assertions=`current candidate is validated and formal evidence is clear`\n\n## Assumptions/Defaults\n- 示例仅用于回放验证：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-flow-lock-direct-local-merge-skill.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"$local-merge-flow"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Direct Local-merge Skill Lock\n\n## Summary\n- 结论：direct flow-skill 引用优先于默认 fast-track。\n\n## Flow Decision\n- `flow_type`: `local-merge`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 走本地合主干规划路径\n  当前请求直接引用 `$local-merge-flow`，因此应绕过默认 fast-track，转为本地 merge-ready / Step 5L Ready 规划。\n\n## Non-goals\n- 不创建 PR 路径的 merge-ready 收口。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 准备本地实现、验证与 merge-ready 证明 (skill: $review-agent + $local-merge-flow)\n\n## Acceptance & Validation\n- direct flow-skill 引用会锁定对应 flow，而不会被 `default_fast_track` 覆盖。\n\n## Assumptions/Defaults\n- 当前消息把 `$local-merge-flow` 作为显式 flow 覆盖：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-flow-lock-fast-flow-keyed.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"flow_type: fast-flow"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Fast Flow Keyed Lock\n\n## Summary\n- 结论：用于验证 keyed flow_type 支持 fast-flow 别名。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary, key implementation detail, and success signal before execution arrangement.\n\n## Non-goals\n- Do not expand beyond the demo contract.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => execute fast-track local validation tasks (skill: $fast-flow)\n\n## Acceptance & Validation\n- Demo acceptance for contract replay.\n\n## Assumptions/Defaults\n- keyed flow_type 别名已解析：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-flow-lock-latest-explicit-user.jsonl

@@ -0,0 +1,3 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"flow_type: normal"}]}}
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"改为 fast-track"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Latest Explicit Flow Lock\n\n## Summary\n- 结论：用于验证用户后续显式改 flow 时，以最新锁定值为准。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary, key implementation detail, and success signal before execution arrangement.\n\n## Non-goals\n- Do not expand beyond the demo contract.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => execute fast-track local validation tasks (skill: $fast-flow)\n\n## Acceptance & Validation\n- Demo acceptance for contract replay.\n\n## Assumptions/Defaults\n- 流程锁定值以最新显式声明为准：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-flow-lock-rui.jsonl

@@ -0,0 +1,3 @@
+{"type":"response_item","payload":{"type":"function_call","name":"request_user_input","call_id":"call_flow_type_lock","arguments":"{\"questions\":[{\"header\":\"流程类型\",\"id\":\"flow_type\",\"question\":\"当前 flow 信号存在歧义，请明确本次流程类型\",\"options\":[{\"label\":\"快车道 (Recommended)\",\"description\":\"推进到 PR merge-ready / Step 5C Ready\"},{\"label\":\"仅实现\",\"description\":\"只产出实现规格\"},{\"label\":\"普通流程\",\"description\":\"验证后本地提交并停下\"},{\"label\":\"本地合主干\",\"description\":\"规划到本地 merge-ready 与本地合并收尾。\"}]}]}"}}
+{"type":"response_item","payload":{"type":"function_call_output","call_id":"call_flow_type_lock","output":"{\"answers\":{\"flow_type\":{\"answers\":[\"仅实现\"]}}}"}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Positive Flow Lock via request_user_input\n\n## Summary\n- 结论：flow_type 已通过 request_user_input 锁定。\n\n## Flow Decision\n- `flow_type`: `implement-only`\n- `flow_source`: `request_user_input`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary, key implementation detail, and success signal before execution arrangement.\n\n## Non-goals\n- Do not expand beyond the demo contract.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => deliver implementation-only spec (skill: $implement-only-flow)\n\n## Acceptance & Validation\n- Demo acceptance for contract replay.\n\n## Assumptions/Defaults\n- 流程类型已锁定：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-flow-lock-standalone-normal-punctuation.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"`normal`."}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Demo Standalone Normal Token With Punctuation\n\n## Summary\n- 结论：standalone flow token 即使带句尾标点，也应优先锁定 `normal`。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 保持 standalone token 的显式覆盖优先级\n  当前消息用 standalone `normal` token 指定 flow，因此不应被 `default_fast_track` 覆盖。\n\n## Non-goals\n- 不回退成默认快车道。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 规划 normal-flow 的实现与本地提交流程 (skill: $normal-flow)\n\n## Acceptance & Validation\n- checker 必须接受带句尾标点的 standalone flow token，并保持 `Flow Decision.flow_type=normal`。\n\n## Assumptions/Defaults\n- 当前消息用 standalone token 显式指定 flow：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-formal-review-sidecar.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请按 normal 规划只读分析 sidecar。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Read Only Analysis Sidecar\n\n## Summary\n- 结论：普通只读分析 sidecar 仍可在计划中声明。\n\n## Flow Decision\n- `flow_type`: `normal`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- G1: 收集实现分析。\n  sidecar 只检查当前实现并返回观察。\n\n## Non-goals\n- N1: 不启动 formal review lane。\n\n## Task Orchestration\n- wave: 1\n  - sub-agent:analysis [gpt-5.6-luna, medium] => 只读检查当前实现和 commit history，返回分析观察 (skill: $normal-flow)\n\n## Acceptance & Validation\n- A1 [contract/direct] sidecar | evidence=`test:analysis` | assertions=`read-only result`\n\n## Assumptions/Defaults\n- sidecar profile 已确定\n</proposed_plan>"}]} }

package/toolkit/skills/dev-plan/references/replay-cases/positive-implement-only-add-tests-reference.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "锁定 implement-only"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Demo Positive Implement-only Add Tests Reference\n\n## Summary\n- 结论：implement-only 允许规划“基于现有 spec 的测试任务”。\n\n## Flow Decision\n- `flow_type`: `implement-only`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary, key implementation detail, and success signal before execution arrangement.\n\n## Non-goals\n- Do not expand beyond the demo contract.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => add integration tests against existing docs/specs/ab123-demo/SPEC.md for parser behavior (skill: $implement-only-flow)\n\n## Acceptance & Validation\n- Demo acceptance for contract replay.\n\n## Assumptions/Defaults\n- 示例仅用于回放验证：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-implement-only-parser-reference.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "锁定 implement-only"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Demo Positive Implement-only Parser Reference\n\n## Summary\n- 结论：implement-only 允许实现任务引用 existing spec 作为输入。\n\n## Flow Decision\n- `flow_type`: `implement-only`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary, key implementation detail, and success signal before execution arrangement.\n\n## Non-goals\n- Do not expand beyond the demo contract.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => add new parser using existing docs/specs/ab123/SPEC.md for validation (skill: $implement-only-flow)\n\n## Acceptance & Validation\n- Demo acceptance for contract replay.\n\n## Assumptions/Defaults\n- 示例仅用于回放验证：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-implement-only-pure.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "锁定 implement-only"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# 仅实现示例计划\n\n## Summary\n- 结论：implement-only 编排仅包含实现规格规划任务。\n\n## Flow Decision\n- `flow_type`: `implement-only`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- 目标 1: 锁定示例范围合同\n  在执行编排前明确结果边界、关键实现细节与成功信号。\n\n## Non-goals\n- 不扩展示例合同之外的范围。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 定义实现范围与接口/数据流边界 (skill: $implement-only-flow)\n- wave: 2\n  - main-agent => 定义失败场景与可验证验收标准 (skill: $implement-only-flow)\n\n## Acceptance & Validation\n- 示例验收用于合同回放验证。\n\n## Assumptions/Defaults\n- 示例仅用于回放验证：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-implement-only-update-parser-reference.jsonl

@@ -0,0 +1,2 @@
+{"type": "response_item", "payload": {"type": "message", "role": "user", "content": [{"type": "input_text", "text": "锁定 implement-only"}]}}
+{"type": "response_item", "payload": {"type": "message", "role": "assistant", "content": [{"type": "output_text", "text": "<proposed_plan>\n# Demo Positive Implement-only Update Parser Reference\n\n## Summary\n- 结论：implement-only 允许更新实现逻辑并引用 existing spec。\n\n## Flow Decision\n- `flow_type`: `implement-only`\n- `flow_source`: `user_explicit`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n\n## Goals\n- Goal 1: lock the demo scope contract\n  Capture the result boundary, key implementation detail, and success signal before execution arrangement.\n\n## Non-goals\n- Do not expand beyond the demo contract.\n\n## Task Orchestration\n- wave: 1\n  - main-agent => update parser behavior with existing docs/specs/ab123/SPEC.md for runtime consistency (skill: $implement-only-flow)\n\n## Acceptance & Validation\n- Demo acceptance for contract replay.\n\n## Assumptions/Defaults\n- 示例仅用于回放验证：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-large-coupled-default-fast-track.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"请规划一次大型重构。所有部分共享同一个验收结果，不能独立验收或独立合并，也不需要 Wave、隔离 Worktree、integration branch 或 aggregate PR。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Coupled Refactor\n\n## Summary\n- 结论：规模较大但不可拆成独立 Ticket，仍按 direct fast-track 收敛。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n- 收口条件：`merge-ready / Step 5C Ready`\n\n## Goals\n- 目标 1: 原子完成耦合重构\n  在一个 topic branch 中完成共享契约、实现与验证。\n\n## Non-goals\n- 不创建多 Ticket Initiative。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 完成实现、验证与 PR 收敛 (skill: $fast-flow)\n\n## Acceptance & Validation\n- 共享验收集通过，PR 达到 Step 5C Ready。\n\n## Assumptions/Defaults\n- 工作不可独立验收或合并：已确定\n</proposed_plan>"}]}}

package/toolkit/skills/dev-plan/references/replay-cases/positive-legacy-direct-no-topology.jsonl

@@ -0,0 +1,2 @@
+{"type":"response_item","payload":{"type":"message","role":"user","content":[{"type":"input_text","text":"恢复一份历史 fast-track 计划。"}]}}
+{"type":"response_item","payload":{"type":"message","role":"assistant","content":[{"type":"output_text","text":"<proposed_plan>\n# Historical Direct Plan\n\n## Summary\n- 结论：恢复缺少 topology 的历史 direct 计划，并停在 merge-ready。\n\n## Flow Decision\n- `flow_type`: `fast-track`\n- `flow_source`: `default_fast_track`\n- `lock_status`: `locked`\n- 开发基线：`origin/main`\n- 收口条件：`merge-ready / Step 5C Ready`\n\n## Goals\n- 目标 1: 恢复历史 direct 交付\n  历史计划没有 topology 字段，恢复时明确按 direct 语义继续。\n\n## Non-goals\n- 不把历史计划转换成 aggregate-stack。\n\n## Task Orchestration\n- wave: 1\n  - main-agent => 恢复历史 direct 实现与 PR 收敛 (skill: $fast-flow)\n\n## Acceptance & Validation\n- 明确标记 legacy-direct 时，缺少 topology 的历史计划按 direct 恢复。\n\n## Assumptions/Defaults\n- 该计划创建于 topology 合同之前：已确定\n</proposed_plan>"}]}}