policy: scope seed-boundary slicing to delegation children

Review fix (ds-review-bot on #623): the unconditional slice regressed the
public SessionStore.fork path — a generic fork child gets seedLength but no
policy baseline, so slicing discarded its seed-carried sandbox/approval
switches and silently widened it to the deployment defaults (a forked
read-only/never parent produced a workspace-write/ask child).

overrideOf now branches on baseline presence: with a header baseline (a
delegation child) the fold covers only own post-seed switches — the
baseline captured from the parent's FULL log subsumes seed history; without
one, the whole log — seeded switches ARE the replayed inherited truth. The
permission preset fold scopes the same way. Red-first: generic-fork
seed-carried override tests in both policy suites.

.agents/notes/implemented/feature/2026-07-25-subagent-policy-inheritance.i18n.yaml

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write
-2026-07-25-subagent-policy-inheritance.md: 8609fee90f3f9ea73f5e6229a38cde5e61337144
-2026-07-25-subagent-policy-inheritance.zh.md: ad1b4650348154d82473b17d920dff475833f00e
+2026-07-25-subagent-policy-inheritance.md: ea8e4626a1bd2ac31c2a7ce207bf9ba73e2e32fc
+2026-07-25-subagent-policy-inheritance.zh.md: b9e978761af7de6ecf9a2e6ec167e2bb512c9ebf

.agents/notes/implemented/feature/2026-07-25-subagent-policy-inheritance.md

@@ -14,7 +14,7 @@ The shared in-process driver (`startInProcessRun` in `packages/subagent/subagent
 
 - **Capture synchronously at delegation, persist in the creation-time header.** The driver reads `overrideOf(parent.session)` for both knobs BEFORE its first await — the delegation moment is the snapshot point, so a parent switch racing the asynchronous child creation belongs to the parent's future, not the child — and stamps the captured values into the child's creation `meta` (`sandboxMode`/`approvalPolicy` on `SessionHeader`). The baseline is durable from the moment the session exists: no listener ordering can starve it (a denying UserPromptSubmit hook that vetoes the first prompt changes nothing), and no crash window can lose it — the decisive case being an idle SessionStart-style injection that persists a complete turn before any prompt turn opens, after which a first-turn event would not yet exist while the session already looks resumable.
 - **Only the override chain is copied, and the owners validate on read.** `overrideOf(session)` — the pure `sandboxOverrideOf`/`approvalOverrideOf` exports, surfaced as service methods — resolves `fold(events past the seed boundary) ?? header baseline`, never the deployment/configured default: an unswitched parent writes no baseline, so its children keep following the LIVE default across restarts. The header fields are neutral strings at the session boundary; each policy owner validates against its closed vocabulary UNCONDITIONALLY on every read (a corrupt header fails loud even when an own switch would shadow it) and throws on foreign values. EVERY knob consumer resolves through the same chain — enforcement (`resolve()`, pty-local) and the permission presets (`current`/`set`), so a child inheriting a wider baseline gets real knob switches when a narrower preset is selected instead of a silent no-op. The driver consumes both services opportunistically (`ctx.get`, type-only imports, `peerDependenciesMeta.optional`): compositions without them delegate policy-free, unchanged.
-- **Fork stale-seed precedence falls out of the seed boundary.** A fork seed may carry the parent's OLD switch events; `overrideOf` folds only events past `header.seedLength`, so seed-carried history is subsumed by the delegation-time baseline while a switch the child makes ITSELF still outranks it. The log stays free of synthetic events — the header is the baseline's one home, and the canonical `setSandboxMode`/`setApprovalPolicy` write paths remain reserved for real runtime switches.
+- **Fork stale-seed precedence falls out of the seed boundary — scoped to delegation children.** A fork seed may carry the parent's OLD switch events; when a header baseline exists, `overrideOf` folds only events past `header.seedLength` — the baseline was captured from the parent's FULL log at delegation, so seed-carried history is subsumed by it while a switch the child makes ITSELF still outranks it. Without a baseline (a top-level session, or a generic `SessionStore.fork` child that captured no policy meta) the fold covers the whole log: there, seeded switches ARE the replayed inherited truth, and slicing them away would silently widen the child to the deployment default. The log stays free of synthetic events — the header is the baseline's one home, and the canonical `setSandboxMode`/`setApprovalPolicy` write paths remain reserved for real runtime switches.
 - **Nesting composes by construction.** A grandchild's capture resolves its parent-the-child's chain (own fold ?? baseline), so the chain collapses one level per delegation, at any depth. One-shot `allowed-once` escalation grants never enter a log or header, so they can never leak down the chain.
 
 ### What a blocked child experiences

.agents/notes/implemented/feature/2026-07-25-subagent-policy-inheritance.zh.md

@@ -14,7 +14,7 @@ Status: implemented
 
 - **委派时同步捕获，持久化在创建时的会话头中。**驱动器在自己的第一个 await 之前就为两个策略旋钮读取 `overrideOf(parent.session)`——委派时刻即快照点，因此与异步的子 agent 创建过程赛跑的父级切换属于父级的未来，而非子 agent——并把捕获值盖章写入子 agent 的创建 `meta`（`SessionHeader` 上的 `sandboxMode`/`approvalPolicy`）。该基线从会话存在的那一刻起就具备持久性：任何监听器顺序都不可能饿死它（即便一个作出拒绝的 UserPromptSubmit 钩子否决了第一个提示词，也不会产生任何影响），任何崩溃窗口也不可能丢失它——决定性的场景是空闲时的 SessionStart 式注入在任何提示词轮次开启之前就持久化了一个完整轮次，在那之后第一个轮次内的事件尚不存在，而会话已经看起来可以恢复。
 - **只复制覆盖链，且由策略 owner 在读取时校验。**`overrideOf(session)`——即纯函数导出 `sandboxOverrideOf`/`approvalOverrideOf`，以服务方法的形式暴露——解析为 `fold(events past the seed boundary) ?? header baseline`，从不包含部署／配置默认值：未切换过的父级不写入任何基线，因此其子 agent 跨重启继续跟随实时默认值。这两个会话头字段在会话边界上只是中性字符串；每个策略 owner 在每次读取时都无条件按自己的封闭词汇校验（即便自己做出的切换会遮蔽基线，损坏的会话头也会大声失败），遇到词汇之外的值即抛出异常。每一个旋钮消费方都经由同一条链解析——强制执行侧（`resolve()`、pty-local）与权限 preset（`current`/`set`）皆然——因此当选中更窄的 preset 时，继承了更宽基线的子 agent 得到的是真实的旋钮切换，而非静默的空操作。驱动器以可选方式消费这两个服务（`ctx.get`，仅类型导入，`peerDependenciesMeta.optional`）：未挂载它们的组合照旧进行无策略委派，行为不变。
-- **fork 陈旧种子的优先级由种子边界自然得出。**fork 种子可能携带父级旧的切换事件；`overrideOf` 只折叠 `header.seedLength` 之后的事件，因此种子携带的历史已被委派时的基线所涵盖，而子 agent 自己做出的切换仍然优先于基线。日志中不含任何合成事件——会话头是基线的唯一存放处，规范写入路径 `setSandboxMode`/`setApprovalPolicy` 仍然只留给真实的运行时切换。
+- **fork 陈旧种子的优先级由种子边界自然得出——仅限委派子 agent。**fork 种子可能携带父级旧的切换事件；当会话头基线存在时，`overrideOf` 只折叠 `header.seedLength` 之后的事件——基线是在委派时从父级的完整日志捕获的，因此种子携带的历史已被它所涵盖，而子 agent 自己做出的切换仍然优先于基线。没有基线时（顶层会话，或未捕获任何策略元数据的通用 `SessionStore.fork` 子会话），折叠覆盖完整日志：此时种子携带的切换本身就是回放所得的继承事实，把它们切掉会把子会话静默放宽到部署默认值。日志中不含任何合成事件——会话头是基线的唯一存放处，规范写入路径 `setSandboxMode`/`setApprovalPolicy` 仍然只留给真实的运行时切换。
 - **嵌套按构造即可组合。**孙代 agent 捕获时解析的是其父级（即上一层的子 agent）的覆盖链（自身折叠 ?? 基线），这条链在每层委派处收拢一级，任意深度均成立。一次性的 `allowed-once` 升级授权从不进入任何日志或会话头，因此永远不可能沿链向下泄漏。
 
 ### 被拦住的子 agent 会经历什么