refactor(status): compose the status task graph in TypeScript

docs/architecture/rfcs/ledger/typescript-control-plane-migration-v0/2026-11-02-status-task-graph.md

@@ -0,0 +1,153 @@
+# Status task-graph composition (2026-11-02)
+
+## Owner boundary
+
+This checkpoint moves the complete optional status task-graph decision from Python
+`work_items/task_graph.py` into the existing TypeScript `work_items/task_graph.ts`.
+One `work_item.task_graph.status` operation owns selected work, claim display,
+user-gate selection/count/cap/overflow, bounded predecessor traversal, run
+selection and lineage joins, evidence, handoff, repair, deduplication, dangling
+edge rejection, timestamps and completeness/limits. It calls existing topology,
+Todo summary, classification and handoff owners in process. The standalone
+`work_item.task_graph.topology` handler is retired; topology remains an internal
+composition function used by status and its semantic tests.
+
+Python supplies source-shaped status facts, latest-run facts and fallback time.
+It retains public-text redaction/normalization facts. Status SHA-256 node/edge
+identity materialization moves into the composer with byte-identical Unicode
+and `None` encoding, so deduplication compares actual historical identities
+rather than unevaluated descriptors. Goal-map uses the same native identity,
+state and topology-row functions; the three Python rule helpers are deleted.
+Source fields are restricted to existing Todo read fields and the consumed
+status lanes. Derived display text uses the qualified composition-only
+compaction boundary described below. There is no Python selection, traversal, gate, evidence,
+handoff, lineage, repair or limit fallback.
+
+`build_goal_task_map` retains its public signature and byte-identical output for
+`chat_status_api.py`. It sends source/public-text facts to `projectGoalTaskMap`,
+which now materializes node/edge identities, state and metadata. Native topology
+fact requests retain their existing compatibility contract. Omitted opaque
+metadata is not stringified; unsupported non-JSON public-text values are marked
+unavailable at the codec edge and rejected only if consumed. Uncalled status
+builders, attachers, source collectors, ref helpers and constants are deleted.
+No new Python/TypeScript basename pair is added.
+
+Physical product diff: +865/−1030, net −165 lines against the archived base.
+`task_graph.py` falls from 1,044 to 255 lines: one module-level recursive public
+text codec and two source/transport facades, with no graph decision rules.
+Tests, fixtures and documentation are counted separately.
+
+## Public text and remaining duplication
+
+Choice (a): there is no second TS redaction rule. Source strings must have
+Python public-text facts; a missing fact throws. The local fallback only calls
+`compactTodoReadText` on registered public fact outputs or explicitly composed
+strings. Interpolated components are asserted to be registered public strings
+or numeric counts. Native tests enumerate composed titles, identities,
+references and reasons, and reject unqualified source text. Constructors are
+claim/lease, evidence prefix and public evidence prefix-slice, handoff identity/
+title/reasons, missing-ID gate, gate overflow, latest-run title, repair defaults,
+and joins of generated node identities. Fixed reason prose comes from native
+graph construction; generated identities contain only ASCII identifier/hash
+characters. Goal scope is already public-normalized by the status caller.
+No `INTERNAL_STATE_SHAPE_PATTERNS` redaction copy remains in this module.
+
+`rg` found no equivalent TS owner of Python `open_todo_items`: `read_codec.ts`
+provides compaction/fields, `context_projection.ts` provides planning/display
+sources, and `summary_lanes.ts` selects normalized summary lanes, not the
+historical source-key, dedup-before-sort, bounded-open display selection.
+Python still serves status attention/monitor projections, project-asset
+readbacks and Todo summary renderers. `statusOpenItems` therefore retains that
+bounded duplication. Exit: migrate those callers to one native Todo open-display
+selection owner, qualify ordering/handoff/limit parity, then remove the Python
+selection and delegate the graph to that owner. Do not substitute normalized
+summary-lane selection with different admission semantics.
+
+The codec caches repeated source strings and computes only field-used caps;
+no source redaction facts are inferred from truncated text. Display caps and the
+unsupported-text marker are local transport facts, not a new shared decision
+vocabulary. Counts below include
+the initial Goal public-text call; “before” means the reviewed encoder draft,
+which called the public codec six times per source string (full safety plus
+five display caps), not the original all-Python graph:
+
+| Status case | Public-text calls before → after |
+| --- | --- |
+| `diamond_at_cap` | 127 → 32 |
+| `cycle` | 79 → 22 |
+| `missing_prerequisite` | 31 → 12 |
+| `open_predecessor` | 79 → 24 |
+| `condition_and_lineage` | 55 → 19 |
+| `predecessor_overflow` | 193 → 47 |
+| `duplicate_identity` | 97 → 19 |
+| `long_public_text` | 61 → 21 |
+| `redaction` | 61 → 21 |
+| `handoff_evidence` | 73 → 21 |
+| `unicode_ids` | 43 → 15 |
+| `all_empty` | 1 → 1 |
+| `gate_overflow` | 145 → 24 |
+| `no_selected_run_gate` | 43 → 17 |
+| `repair_evidence` | 79 → 29 |
+| `stale_source_time` | 43 → 18 |
+| `absent_goal` | 1 → 1 |
+| `invalid_goal` | 1 → 1 |
+
+Largest characterization request (`long_public_text`): compact JSON parameters
+with the transport's ASCII escaping, 6,797 → 4,899 bytes (−27.9%); calls
+61 → 21 (−65.6%). These are parameter bytes, excluding RPC framing. The maximum
+call case, predecessor overflow, falls from 193 to 47. This is codec/request
+cost evidence, not an end-to-end latency or effectiveness claim.
+
+## Crossings and parity
+
+Instrumentation counts every `effect_runtime_result` reached by the public
+Python graph entrypoint, including Todo display/handoff helpers:
+
+| Operation | Before → after |
+| --- | --- |
+| Selected work, including topology/error evidence cases | 4 → 1 |
+| User-gate overflow | 7 → 1 |
+| Run/gate nodes with no selected Todo | 4 → 1 |
+| Valid Goal with all-empty facts and no graph | 0 → 1 |
+| Absent or invalid Goal | 0 → 0 |
+| Goal task map | 1 → 1 |
+
+The empty valid snapshot is decided inside the native composer; only absent or
+invalid Goals return before transport. A rejected native request has no Python
+fallback. Large packets reuse the existing local snapshot transport.
+
+The frozen synthetic characterization corpus was green on unchanged
+`fb16c6cb1369efce52a673030c914b38e1dac166` (21 cases) before implementation;
+expected JSON bytes remain unchanged. Independent native cases test semantics
+rather than using that corpus as a specification. The status Markdown renderer
+has no graph section. Real `status --include-task-graph` CLI output matches an
+archived base byte-for-byte on isolated File and SQLite homes in JSON and
+Markdown, with a fixed observation clock: respectively 33,417/33,445 and
+5,120/5,126 bytes. The actual chat `_goal_task_map` entrypoint, using the same
+canonical File/SQLite sources, matches at 2,310 bytes on each provider. The three
+already-pinned Goal-map cases were green on the archived base before refactoring;
+the fixture expected outputs have not been edited.
+
+## Qualification and exit
+
+Current rerun: Python 355 passed, 2 failed across the six requested suites and
+all tests found by `rg build_goal_task_map / chat_status_api`; native graph/handler/
+loading tests 50 passed. Ruff (two changed Python files), control-plane typecheck,
+task-graph fixture, attention queue, semantic vocabulary and maintainability
+smokes, six CLI/chat parity comparisons and `git diff --check` passed. Earlier
+interface-budget/integrated-canary evidence remains historical, not a new rerun. The quota planning-horizon suite retains two failures
+(missing `planning_horizon`, File and SQLite), reproduced with the same tests
+in an untouched `git archive HEAD` baseline copy; they are not reported as passes.
+No gate, budget, vocabulary anchor or baseline is relaxed.
+
+The Python status facade exits when status callers deliver the source/public
+text facts directly to the native owner and the retained public-text codec has
+an independently qualified native replacement. Remove its operation only when
+those callers use an in-process owner. The same codec exit applies to
+the Goal-map facade. The characterization corpus
+can retire once facade removal has independent public-entrypoint coverage.
+
+The adjacent cleanup removes the obsolete topology RPC and callback surface;
+no unrelated refactor or product journey is introduced. Existing CLI, attention
+queue, chat and review-packet consumers keep their public payloads. No end-to-end latency,
+total resource-cost or long-horizon effectiveness/efficiency benefit is claimed.

docs/architecture/rfcs/ledger/typescript-control-plane-migration-v0/2026-11-02-status-task-graph.zh-CN.md

@@ -0,0 +1,125 @@
+# Status 任务图组合（2026-11-02）
+
+## Owner 边界
+
+本检查点将可选 status 任务图的完整决策从 Python `work_items/task_graph.py`
+迁入现有 TypeScript `work_items/task_graph.ts`。单个
+`work_item.task_graph.status` 操作拥有所选工作、claim 展示、用户 gate
+选择/计数/上限/溢出、有界前驱遍历、run 选择与 lineage 连接、证据、handoff、
+repair、去重、悬空边拒绝、时间戳及完整性/限制。它在进程内组合现有 topology、
+Todo 摘要、分类和 handoff owner。独立的 `work_item.task_graph.topology`
+handler 已退役；topology 作为内部组合函数保留，由 status 及其语义测试使用。
+
+Python 提供源形状的 status 事实、latest-run 事实和后备时间。它保留公开文本
+脱敏/规范化事实。status 的 SHA-256 节点/边身份物化迁入组合器，Unicode 和
+`None` 编码逐字节一致，从而使用实际历史身份去重，而非尚未求值的 descriptor。
+Goal-map 也复用同一套原生身份、状态及 topology-row 函数；三个 Python 规则 helper 已删除。源字段限定为现有 Todo 读取字段及实际消费的 status lane。
+派生展示文本使用下文限定的组合文本压缩边界。不存在 Python 的选择、遍历、
+gate、证据、handoff、lineage、repair 或限制后备决策。
+
+`build_goal_task_map` 保持公开签名及 `chat_status_api.py` 输出逐字节一致。
+它向 `projectGoalTaskMap` 提交源数据/公开文本事实，由原生实现物化节点/边身份、
+状态及元数据。现有原生 topology 事实请求保留兼容合同。被省略的 opaque 元数据
+不会提前字符串化；非 JSON 公开文本值在 codec 边界标记为不可用，仅在消费时拒绝。
+无调用方的 status builder、attacher、源收集器、ref helper 和常量已删除。
+没有新增 Python/TypeScript 同名文件对。
+
+相对归档基线，产品物理行差异为 +865/−1030，净 −165 行。
+`task_graph.py` 从 1,044 行降至 255 行：一个模块级递归公开文本 codec 和两个
+源数据/传输 facade，不含任务图决策规则。测试、fixture 和文档单独计数。
+
+## 公开文本与保留的重复规则
+
+选择 (a)：没有第二套 TS 脱敏规则。源字符串必须提供 Python 公开文本事实，
+缺少事实时抛错。本地 fallback 仅对已登记的公开事实输出或显式组合字符串调用
+`compactTodoReadText`。插值组件断言为已登记公开字符串或数字计数。原生测试枚举
+组合标题、身份、引用及原因，并拒绝未限定的源字符串。构造器包括 claim/lease、
+evidence 前缀及公开证据的前缀切片、handoff 身份/标题/原因、缺失 ID 的 gate、
+gate 溢出、latest-run 标题、repair 默认文本和生成节点身份的连接。固定原因来自
+原生图构造；生成身份仅含 ASCII 标识符/hash 字符。status 调用方已对 Goal scope
+做公开规范化。模块不再复制 `INTERNAL_STATE_SHAPE_PATTERNS` 脱敏规则。
+
+`rg` 未找到 Python `open_todo_items` 的等价 TS owner：`read_codec.ts` 提供
+压缩/字段，`context_projection.ts` 提供规划/展示源，`summary_lanes.ts` 选择
+规范化摘要 lane，不实现历史 source-key、先去重再排序、有界开放项展示选择。
+Python 仍服务 status attention/monitor、project-asset 读回及 Todo 摘要渲染。
+因此 `statusOpenItems` 暂时保留这项有限重复。退出条件：这些调用方迁入统一的
+原生 Todo open-display 选择 owner，验收排序/handoff/上限一致性，再删除 Python
+选择并让图委托同一 owner。不得用准入语义不同的摘要 lane 选择替代。
+
+codec 缓存重复源字符串，只计算字段实际使用的上限；不从截断文本推断源脱敏事实。展示上限和不支持文本标记仅是本地传输事实，
+不构成新的共享决策词汇。
+下表包含初始 Goal 公开文本调用；“之前”指本次评审的 encoder 初稿，每个源字符串
+调用六次（完整安全检查加五个展示上限），并非最初的全 Python 图：
+
+| Status 用例 | 公开文本调用次数：之前 → 之后 |
+| --- | --- |
+| `diamond_at_cap` | 127 → 32 |
+| `cycle` | 79 → 22 |
+| `missing_prerequisite` | 31 → 12 |
+| `open_predecessor` | 79 → 24 |
+| `condition_and_lineage` | 55 → 19 |
+| `predecessor_overflow` | 193 → 47 |
+| `duplicate_identity` | 97 → 19 |
+| `long_public_text` | 61 → 21 |
+| `redaction` | 61 → 21 |
+| `handoff_evidence` | 73 → 21 |
+| `unicode_ids` | 43 → 15 |
+| `all_empty` | 1 → 1 |
+| `gate_overflow` | 145 → 24 |
+| `no_selected_run_gate` | 43 → 17 |
+| `repair_evidence` | 79 → 29 |
+| `stale_source_time` | 43 → 18 |
+| `absent_goal` | 1 → 1 |
+| `invalid_goal` | 1 → 1 |
+
+最大刻画请求 `long_public_text` 使用传输层 ASCII 转义的紧凑 JSON 参数计量，
+6,797 → 4,899 字节（−27.9%），调用 61 → 21（−65.6%）。这些是参数字节，
+不含 RPC framing。最大调用用例 predecessor overflow 从 193 降至 47。
+这是 codec/请求成本证据，不代表端到端延迟或效果收益。
+
+## 跨运行时调用与一致性
+
+插桩统计公开 Python 图入口触发的每次 `effect_runtime_result`，包括 Todo
+展示/handoff helper：
+
+| 操作 | 迁移前 → 后 |
+| --- | --- |
+| 所选工作，包括 topology/错误证据情形 | 4 → 1 |
+| 用户 gate 溢出 | 7 → 1 |
+| 无所选 Todo、但有 run/gate 节点 | 4 → 1 |
+| 有效 Goal、事实全空且无图 | 0 → 1 |
+| 缺失或无效 Goal | 0 → 0 |
+| Goal 任务地图 | 1 → 1 |
+
+有效空快照由原生组合器决策；只有缺失或无效 Goal 在传输前返回。原生请求拒绝后
+没有 Python fallback。大请求复用现有本地快照传输。
+
+冻结的合成刻画语料在实现前已于未修改的
+`fb16c6cb1369efce52a673030c914b38e1dac166` 通过（21 个用例）；预期 JSON
+字节保持不变。独立原生用例验证语义，而不将刻画语料当作规范。status Markdown
+渲染器没有图专属章节。真实 `status --include-task-graph` CLI 在隔离的 File 和
+SQLite home 上，以固定观察时钟对比归档基线，JSON 和 Markdown 均逐字节相同：
+分别为 33,417/33,445 及 5,120/5,126 字节。实际 chat `_goal_task_map`
+入口使用相同的规范 File/SQLite 源，两者均逐字节一致，为 2,310 字节。三个已冻结的
+Goal-map 用例在重构前已于归档基线通过；fixture 预期输出未修改。
+
+## 验收与退出
+
+本次重跑：六个指定 Python 套件及 `rg build_goal_task_map / chat_status_api`
+找到的全部测试共 355 通过、2 失败；原生图/handler/loading 测试 50 通过。
+两个变更 Python 文件的 Ruff、control-plane typecheck、任务图 fixture、attention
+queue、语义词汇、可维护性 smoke、六项 CLI/chat 一致性和 `git diff --check`
+均通过。早先接口预算和集成 canary 证据仅保留为历史，不声称本次重跑。
+quota planning-horizon 套件仍有两项
+失败（File 和 SQLite 缺少 `planning_horizon`），使用同样测试在未修改的
+`git archive HEAD` 基线副本中复现；不将其报告为通过。未放宽任何 gate、预算、
+词汇锚点或基线。
+
+当 status 调用方直接向原生 owner 提供源/公开文本事实，且保留的公开文本 codec
+已有独立验收的原生替代时，Python status facade 才可退出。只有这些调用方
+改为进程内 owner 后，才移除其操作。Goal-map facade 使用相同的 codec 退出条件。facade 移除具备独立公开入口覆盖后，可退役刻画语料。
+
+相邻清理移除了过时 topology RPC 和 callback 表面；没有引入无关重构或产品
+旅程。现有 CLI、attention queue、chat 及 review-packet 消费方保持其公开
+payload。不声称端到端延迟、完整资源成本或长程效果/效率收益。

docs/architecture/rfcs/typescript-control-plane-migration-v0.md

@@ -2843,6 +2843,7 @@ with shared write/read status-phase checks. See the [2026-10-02 checkpoint](ledg
 - Complete global summary and gate read transactions: [2026-11-02 checkpoint](ledger/typescript-control-plane-migration-v0/2026-11-02-global-manager-reads.md).
 - Complete global-risk read ownership: [2026-11-02](ledger/typescript-control-plane-migration-v0/2026-11-02-global-risks-read.md).
 - Complete durable-completion evidence read: [2026-11-02 owner and crossings ledger](ledger/typescript-control-plane-migration-v0/2026-11-02-completion-evidence-read.md).
+- Complete status task-graph composition: [2026-11-02 checkpoint](ledger/typescript-control-plane-migration-v0/2026-11-02-status-task-graph.md).
 - Complete adaptive child admission decision: [2026-11-02 checkpoint](ledger/typescript-control-plane-migration-v0/2026-11-02-adaptive-child-admission.md).
 - Progress-review receipt qualification: [2026-11-02 checkpoint](ledger/typescript-control-plane-migration-v0/2026-11-02-progress-review-qualification.md).
 - Global Todo readiness complete owner cutover (both consumers): [2026-11-02 ledger](ledger/typescript-control-plane-migration-v0/2026-11-02-global-todo-readiness.md).

docs/architecture/rfcs/typescript-control-plane-migration-v0.zh-CN.md

@@ -2108,6 +2108,7 @@ SQLite 中位数略升，收益不能外推为全面加速。这只资格化本
 - 完整全局摘要与门禁读取事务：[2026-11-02 检查点](ledger/typescript-control-plane-migration-v0/2026-11-02-global-manager-reads.zh-CN.md)。
 - 完整全局风险读取所有权：[2026-11-02](ledger/typescript-control-plane-migration-v0/2026-11-02-global-risks-read.zh-CN.md)。
 - 完整 durable-completion 证据读取：[2026-11-02 owner 与调用边界记录](ledger/typescript-control-plane-migration-v0/2026-11-02-completion-evidence-read.zh-CN.md)。
+- 完整 status 任务图组合：[2026-11-02 检查点](ledger/typescript-control-plane-migration-v0/2026-11-02-status-task-graph.zh-CN.md)。
 - 完整自适应子 Agent 准入决策：[2026-11-02 检查点](ledger/typescript-control-plane-migration-v0/2026-11-02-adaptive-child-admission.zh-CN.md)。
 - 进度评审 receipt 资格判定：[2026-11-02 检查点](ledger/typescript-control-plane-migration-v0/2026-11-02-progress-review-qualification.zh-CN.md)。
 - Global Todo 就绪分类完整 owner 切换（两个消费者）：[2026-11-02 记录](ledger/typescript-control-plane-migration-v0/2026-11-02-global-todo-readiness.zh-CN.md)。