@fitlab-ai/agent-infra 0.9.7 → 0.9.8

package/templates/.agents/hooks/lifecycle-delegation.js

@@ -147,11 +147,11 @@ process.stdin.on('end', () => {
     args.push(
       '--native-agent', String(nativeAgent),
       '--child-id', String(event.child_id || event.agent_id || event.thread_id || ''),
-      '--parent-id', String(event.parent_id || event.parent_session_id || event.session_id || ''),
-      '--spawn-mode', String(event.spawn_mode || (client === 'claude-code' ? 'fresh' : ''))
+      '--parent-id', String(event.parent_id || event.parent_session_id || event.session_id || '')
     );
+    if (event.spawn_mode) args.push('--spawn-mode', String(event.spawn_mode));
     if (event.model) args.push('--actual-model', String(event.model));
-    const actualEffort = event.actual_reasoning_effort || event.reasoning_effort || event.model_reasoning_effort;
+    const actualEffort = event.effort?.level || event.actual_reasoning_effort || event.reasoning_effort || event.model_reasoning_effort;
     if (actualEffort) args.push('--actual-reasoning-effort', String(actualEffort));
     if (event.model_fallback_reason) args.push('--model-fallback-reason', String(event.model_fallback_reason));
     if (event.reasoning_effort_fallback_reason) {
@@ -162,6 +162,11 @@ process.stdin.on('end', () => {
       '--native-agent', String(nativeAgent),
       '--child-id', String(event.child_id || event.agent_id || event.thread_id || '')
     );
+    if (client === 'claude-code') {
+      if (event.model) args.push('--actual-model', String(event.model));
+      const actualEffort = event.effort?.level || event.actual_reasoning_effort || event.reasoning_effort;
+      if (actualEffort) args.push('--actual-reasoning-effort', String(actualEffort));
+    }
   }
   try {
     const { command, commandArgs, shell } = internalCliInvocation(args);

package/templates/.agents/rules/lifecycle-orchestration.en.md

@@ -5,7 +5,7 @@
 - The orchestrator only routes and delegates. Stage skills remain the single source of truth for behavior and artifacts.
 - Every stage and rework round uses a fresh executor. Every review uses a fresh reviewer; follow-up reuse is forbidden.
 - A reviewer may only write its review artifact and core-generated task metadata. Business-code, HEAD, or index changes invalidate the receipt.
-- An active run has one pending delegation. The child must pass the activation barrier before any stage side effect. Missing, late, mismatched, forked, replayed, hook-less, or drifted evidence pauses the run.
+- An active run has one pending delegation. The child must pass the activation barrier before any stage side effect. Missing, late, mismatched, forked, replayed, hook-less, or drifted evidence pauses the run (on the claude-code path, missing or mismatched model/effort evidence, and fork/spawn-mode evidence this host structurally does not provide, follow the recording rules in `.agents/skills/run-task/reference/host-validation.md` instead and are exempt from this; missing or mismatched `parentId`/`childId` still pauses).
 - The first release ends after one existing safely gated `commit`; it does not create a PR, monitor checks, or complete the task.
 
 ## Recovery
@@ -22,8 +22,9 @@
 
 - A new run persists model and reasoning effort for both roles. Explicit policy is atomic across all four role fields; only a fully absent explicit policy may fall back to the current client's `agentClients[].orchestration`.
 - Route resolves requested model/effort by role, and prepare matches both before snapshotting. Native spawn must not inherit session defaults.
-- Native start records host-observed actual model/effort. Each mismatch needs its own fallback reason, and requested values must never be fabricated as actual evidence.
+- Native start records host-observed actual model/effort. Each mismatch needs its own fallback reason (the claude-code path may leave this blank and follow the recording rules instead), and requested values must never be fabricated as actual evidence.
 - Model selection is labeled as a complete catalog, partial catalog, or interactive-only guidance; a local override enum must not be presented as complete.
+- claude-code's requested reasoning effort does not yet support per-role dispatch (concurrent tasks would race on a shared file); `delegationEvidence.actualReasoningEffort` is declared `spawn-ack`, meaning it is only recorded when honestly observed in a native spawn lifecycle event (Start/Stop) — it is not a promise of per-role dispatch and does not gate activation.
 
 ## Stable Pause Conditions
 

package/templates/.agents/rules/lifecycle-orchestration.zh-CN.md

@@ -5,7 +5,7 @@
 - 总控只路由和委派；阶段技能仍是业务规则与产物格式的单一事实源。
 - 每个阶段和每轮返工都创建 fresh executor；每轮审查创建 fresh reviewer，禁止 follow-up 复用。
 - reviewer 只能写当前审查产物和核心生成的任务元数据。业务代码、HEAD 或暂存区变化会使 receipt 失效。
-- active run 中只有一个 pending delegation。child 必须在任何阶段副作用前通过 activation barrier；缺失、超时、错配、fork、重放、hook 不可用或工作区漂移一律暂停。
+- active run 中只有一个 pending delegation。child 必须在任何阶段副作用前通过 activation barrier；缺失、超时、错配、fork、重放、hook 不可用或工作区漂移一律暂停（claude-code 路径下 model/effort 证据的缺失或错配、以及该宿主结构上不提供的 fork/spawn-mode 证据，按 `.agents/skills/run-task/reference/host-validation.md` 的记录规则处理，不在此列；`parentId`/`childId` 的缺失或错配仍然暂停）。
 - 首版成功终点是一次通过既有安全门禁的 `commit`；不创建 PR、不监控 checks、不执行 `complete-task`。
 
 ## 恢复语义
@@ -22,8 +22,9 @@
 
 - 新 run 必须固化 executor/reviewer 各自的 model 与 reasoning effort；显式策略必须四字段原子完整，完全没有显式字段时才读取当前 client 的 `agentClients[].orchestration`。重入不得静默改写策略。
 - route 按 role 返回 requested model/effort，prepare 必须在工作区快照前精确匹配两者；原生 spawn 不能继承会话默认值。
-- 原生 start 必须记录宿主观察到的 actual model/effort。任一字段与 requested 不同时必须记录独立 fallback reason；requested 值不得补造 actual 证据。
+- 原生 start 必须记录宿主观察到的 actual model/effort。任一字段与 requested 不同时必须记录独立 fallback reason（claude-code 路径下允许留空，按记录规则处理）；requested 值不得补造 actual 证据。
 - 模型选择能力必须标记 complete catalog、partial catalog 或 interactive-only guidance；局部 override 枚举不得冒充完整目录。
+- claude-code 的 requested reasoning effort 暂不支持按角色下发（跨任务共享文件存在写入竞态）；`delegationEvidence.actualReasoningEffort` 声明为 `spawn-ack`，语义为"仅能在宿主原生 spawn 生命周期事件（Start/Stop）里如实观察到时记录，不构成按角色下发的承诺，也不构成放行门禁"。
 
 ## 稳定暂停条件
 

package/templates/.agents/skills/run-task/reference/host-validation.en.md

@@ -6,7 +6,9 @@ Before publishing support for a client, prove the lifecycle evidence chain with
 
 - Client version, enabled multi-agent capability, launch surface, and trust boundary.
 - Raw start/stop events, or the parent PostTool spawn/completed-wait fallback when custom roles omit them, stably correlate the managed agent, parent/child identity, fresh spawn mode, actual model, and actual reasoning effort.
+  > Direct-host clients that structurally do not emit fork/spawn-mode event fields (currently claude-code) may derive identity correlation solely from parent/child identity uniqueness plus the receipt's one-shot state machine, without an additional fresh-spawn-mode proof; the client must still honestly record this gap and how the substitute guarantee is composed, and must never fabricate spawn-mode evidence. This exception does not apply to any other client. Concretely, claude-code's identity guarantee is provided only by non-empty and mutually distinct `parentId`/`childId`, a unique `childId` within the same run, and the receipt's one-shot state machine — it carries none of Codex's `spawn_mode` fork proof and is weaker than Codex's.
 - Explicit requested model/effort reaches each executor/reviewer. A fallback of either field reports the corresponding actual value and its own non-empty reason.
+  > claude-code exception: requested reasoning effort does not yet support per-role dispatch to executor/reviewer — `.claude/agents/{executor,reviewer}.md` is a repository-level shared singleton, and writing it per role would race across concurrent tasks. This client only needs to honestly record the actual model/actual reasoning effort observed in native Start/Stop host events (`delegationEvidence.actualReasoningEffort` is declared `spawn-ack`); when not observed it is recorded as missing, which is not treated as a fallback requiring its own reason and does not fail closed. This exception does not apply to any other client.
 - Candidate-checkout and packed-install behavior match, with model policy, receipts, and verification results auditable in `orchestration.json`.
 - Direct-host project/managed source and sandbox isolated-user source preserve the same build, contract, profile, and controller provenance through prepare/start/stop/consume.
 
@@ -19,4 +21,4 @@ Before publishing support for a client, prove the lifecycle evidence chain with
 5. Commit only redacted summaries and non-sensitive fixtures. Remove or replace tokens, user-specific absolute paths, transcript content, and credentials.
 6. For sandbox, run at least ten cold executor starts and ten cold reviewer starts. Record monotonic prepared, spawn-dispatch, SubagentStart, and activation-completed timestamps plus p50/p95/max. Enable only when max plus 20% fits the deadline and symlink/config/plugin/lease/context/build failure injection and terminal cleanup audits pass.
 
-If any required field cannot be observed reliably from the real host, keep that client's orchestration capability `unsupported` and record the gap as manual validation. Codex may advertise `experimental`, but every `prepare` must still pass static preflight and either native start/stop or the parent spawn/completed-wait fallback must produce verifiable consumed host evidence. The fallback may wait only for empty turns or protocol `inProgress`; malformed, identity/transport errors, or abnormal terminals pause deterministically.
+If any required field cannot be observed reliably from the real host, keep that client's orchestration capability `unsupported` and record the gap as manual validation. Codex may advertise `experimental`, but every `prepare` must still pass static preflight and either native start/stop or the parent spawn/completed-wait fallback must produce verifiable consumed host evidence. The fallback may wait only for empty turns or protocol `inProgress`; malformed, identity/transport errors, or abnormal terminals pause deterministically. The claude-code exception above is already scoped in the second bullet and does not change this rule's fail-closed requirement for any other field or client.

package/templates/.agents/skills/run-task/reference/host-validation.zh-CN.md

@@ -6,7 +6,9 @@
 
 - 客户端版本、启用的多代理能力、启动入口与信任边界。
 - 原始 start/stop 事件，或 custom role 缺少它们时的 parent PostTool spawn/completed-wait fallback，能稳定关联 managed agent、parent/child identity、fresh spawn mode、actual model 和 actual reasoning effort。
+  > 结构上不提供 fork/spawn-mode 事件字段的 direct-host 客户端（当前为 claude-code），身份关联允许仅由 parent/child identity 的唯一性与一次性状态机提供，不要求额外的 fresh spawn mode 证明；该客户端仍需如实记录这一缺口与替代保证的构成，不得补造 spawn mode 证据。此例外不适用于其他客户端。具体而言，claude-code 的身份保证仅由 `parentId`/`childId` 非空且互不相等、同一 run 内 `childId` 唯一、receipt 一次性状态机三者提供，不含 codex 式的 `spawn_mode` fork 证明，强度弱于 codex。
 - 显式 requested model/effort 能传给 executor/reviewer；宿主对任一字段回退时，事件能给出对应 actual 值与独立的非空 fallback reason。
+  > claude-code 例外：requested reasoning effort 暂不支持按角色下发给 executor/reviewer——`.claude/agents/{executor,reviewer}.md` 是仓库级共享单例，动态写入存在跨任务竞态，解决该竞态超出本任务范围。该客户端只要求如实记录宿主原生 Start/Stop 事件中能观察到的 actual model/actual reasoning effort（`delegationEvidence.actualReasoningEffort` 声明为 `spawn-ack`）；观察不到时记录为缺失，不视为需要独立 fallback reason 的回退场景，也不构成 fail-closed 阻断。此例外不适用于其他客户端。
 - candidate checkout 与打包安装后的行为一致，且模型策略、receipt 与验证结果可从 `orchestration.json` 复核。
 - direct-host 的 project/managed source 与 sandbox controller 的 isolated-user source、build/contract/profile/controller digest 必须在 prepare/start/stop/consume 全链路一致。
 
@@ -19,4 +21,4 @@
 5. 仅把去敏摘要与非敏感 fixture 纳入版本库；token、绝对用户路径、transcript 内容和凭证必须删除或替换。
 6. sandbox backend 另做至少 10 次 executor 与 10 次 reviewer 冷启动，记录 prepared、spawn dispatch、SubagentStart、activation completed 的单调时钟及 p50/p95/max；仅当 max 加 20% 余量不超过 deadline，并通过 symlink/config/plugin/lease/context/build 失败注入和终态清理审计时启用。
 
-任一字段无法从真实宿主稳定观察时，该客户端的 orchestration capability 保持 `unsupported`，并把缺口记录为人工验证项。Codex 可声明为 `experimental`，但每次 `prepare` 仍必须通过静态 preflight，且原生 start/stop 或 parent spawn/completed-wait fallback 必须形成可验证的 consumed host evidence。fallback 只允许 empty turns/协议 `inProgress` 等待；malformed、身份/传输错误或异常 terminal 都稳定暂停。
+任一字段无法从真实宿主稳定观察时，该客户端的 orchestration capability 保持 `unsupported`，并把缺口记录为人工验证项。Codex 可声明为 `experimental`，但每次 `prepare` 仍必须通过静态 preflight，且原生 start/stop 或 parent spawn/completed-wait fallback 必须形成可验证的 consumed host evidence。fallback 只允许 empty turns/协议 `inProgress` 等待；malformed、身份/传输错误或异常 terminal 都稳定暂停。上述关于 claude-code 的例外已在第 2 条单独说明，不改变本条对其余字段/客户端的 fail-closed 要求。