docs(contributing): mark GH-C07 landed and point at its regression

[kokokoXUY]

Goal And Delivered Outcome

• Goal/source and gap: the state entry for GH-C07 still reads as available work — "extend the same lock or optimistic-revision guard to per-goal todo/refresh/history writers and include a concurrent todo add/update regression" — but test(todos): cover concurrent add-update serialization #4048 (merged 2026-09-07) shipped exactly that regression.
• Observable before → after: a contributor reading the board sees an open task and can spend a slice implementing something that already exists (I did); the row now records the landing and the command that proves it.
• Issue/task and intended base: Related to [Task]: GH-C07 characterize concurrent Todo add/update serialization #3893 (the GH-C07 characterization issue, closed by test(todos): cover concurrent add-update serialization #4048). Base main.

Scope And Continuation

• Completed scope and remaining work: the GH-C07 row now reads Landed: #4048 (merged 2026-09-07), names tests/test_todo_write_serialization.py and the shared legacy writer fence it drives, keeps the global-registry coverage reference, and replaces the validation cell with the command that actually proves the dimension.
• Slice boundary / successor: complete within this scope.

Validation

• Tested revision: 84179d12b7ba01739358d609e51fd6b54c2a880a
• Run state: finished
• Input classes: none

Check kind	Result	Public-safe evidence / limitation
unit	passed	pytest -q tests/test_todo_write_serialization.py tests/test_global_registry_write_serialization.py → 8 passed, 1 skipped. That is the command the row now cites.
static	passed	loopx check --scan-path docs/development/contributor-tasks.md → errors=0 with a clean public boundary scan. The two warnings are a missing local registry file on this host and are identical before and after the change.

• Coverage and gaps: a public documentation row only; no runtime path is touched. I confirmed the shipped regression exists and passes on current main before claiming the landing, rather than trusting the issue text — that text still says add_goal_todo and update_goal_todo "now both hold exclusive_file_lock", which no longer describes where the lock lives, so the row names the legacy writer fence instead of the helper.

Frontend / Visual Evidence

Not applicable: contributor documentation only.

[kokokoXUY]

Queue note: #5080 (docs(contributor-board): admit tasks only with a roadmap, RFC, or adoption evidence) rewrites this file (+101 −141) and restructures the rows this PR
edits. The substantive staleness this PR fixes is not in #5080's diff — its GH-C02 row
still reads Claimed: PR #4436 for a pull request that merged on 2026-09-15 — but the two
will conflict textually. If #5080 lands first I will re-apply this change onto the new board
rather than ask anyone to resolve it.

[huangruiteng]

APPROVE：未发现阻塞问题。已独立核对完整差异、合并事实和真实调用链，并在精确 head 84179d12b7ba01739358d609e51fd6b54c2a880a 运行本地验证；本次不执行合并。

动机

GH-C07 旧行仍邀请贡献者增加并发 Todo add/update 回归，但 #4048 已于 2026-09-07 合并。我核对了 GitHub 合并记录、该合并提交在当前主干中的祖先关系，以及主干仍保留的测试和调用边界。因此这不是用“已合并”替代验收，而是消除会诱导重复实现的过期任务描述。本 PR 完成的是这一个公共任务行的纠偏，不宣称整个状态存储或并发体系已完成。

改动思路

沿用现有贡献任务板，把历史落地事实和可重跑证据放回原 GH-C07 行，不新增任务身份或第二份运行时权威。测试调用真实 add_goal_todo / update_goal_todo；当前 legacy_todo_write_transaction 在读取和写回期间持有 per-Goal 协调锁与状态文件锁。刷新、bootstrap、反馈、事件写入等相邻调用点也已核对其持锁边界或明确的调用者锁契约。

正向路径是：贡献者阅读任务行 → 找到合并记录和所属回归 → 用兼容的 checkout 环境验证 → 不再重复领取已交付维度。并发测试里的锁被替换为确定性线程锁；它证明真实 mutation 的持锁次序及最终双写保留，不单独证明所有 OS / provider 的锁实现。

具体改动

关键内容讲解

• docs/development/contributor-tasks.md:165：从仍可领取的“增加并发回归”改为 Landed: #4048，说明其实际交付是已有 add/update 回归，而不是声称本 PR 修改了生产锁实现。
• 同一行保留全局 registry 的独立覆盖，并将验证入口替换为两份所属测试；核对了 atomic_write_state_text 的 sibling-lock 契约及调用边界，避免把原子替换错误等同于 read-modify-write 串行化。
• 完整差异只有一个文档行，+1/-1；没有运行时代码、schema、CLI、自动安装的 Agent 指令或权限变更，也没有改变公共首屏。

对主干的风险

本地在该 head 的 Python 3.13 checkout 环境执行：

• uv run --extra test python -m pytest -q -rs tests/test_todo_write_serialization.py tests/test_global_registry_write_serialization.py：9 passed，含并发双方结果保留、全局锁内读写和真实子进程覆盖。
• uv run --extra test python examples/docs-governance-smoke.py：通过。
• CI 路径分类器确认 core_tests=false；分类器自身 7 项测试、DCO trailer 核对和 git diff --check 均通过。
• 文档 public-boundary scan：1 文件 clean、errors=0。空合成 runtime 的未初始化提示不是文档问题。按当前 review 配置未查询或等待远端 CI。

剩余风险主要是文档集成：#5114 携带完全相同的 GH-C07 行；#5080 在 006d9edcd4fab829722476933011accdb7131bd6 重排任务板时，仍将 GH-C07 写为缺少 per-goal 保护及回归。后者若随后合并，应保留这里已验证的事实，不能恢复旧任务。这是另一候选的集成风险，不是本 diff 的缺陷。

我的整体评价

long_horizon=improved：清除重复工作的入口；user_experience=improved：读者得到落地事实和可执行验证，而非过期邀请。没有改变调度、Todo 机器状态、默认开关、强制义务或 actor 权限，相关运行时审查项不适用。

未来向简化检查已考虑任务板维护边界：直接更新现有行足够，不值得加入自动同步框架或再造状态存储；与 #5080 的布局调整集成时应复用同一任务身份。尚未做跨平台/全 provider 锁资格验证，但本 PR 没有修改这些实现，不能把该未测范围当作阻塞此事实纠偏的理由。

English verdict: APPROVE - The GH-C07 landing and regression references are independently verified; no blocking finding on this exact docs-only head.