intentdna 1.9.0 → 1.9.3

package/.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
     {
       "name": "intentdna",
       "description": "Declarative policy layer for AI agent behavior with plugin-managed hook runtime for Claude Code.",
-      "version": "1.9.0",
+      "version": "1.9.3",
       "author": {
         "name": "Samuel"
       },
@@ -25,5 +25,5 @@
       ]
     }
   ],
-  "version": "1.9.0"
+  "version": "1.9.3"
 }

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "intentdna",
-  "version": "1.9.0",
+  "version": "1.9.3",
   "description": "Declarative policy layer for AI agent behavior",
   "author": {
     "name": "Samuel"

package/README.md

@@ -1,26 +1,35 @@
 # IntentDNA
 
-**把真实、重复、可验证的 AI 工作方法，做成可独立运行的工作流 CLI。**
+**面向 Codex CLI、Claude Code 等 AI 编程环境的声明式工作流资产、同步与运行工具。**
 
 [English](#english) | [路线转换文章](https://wanglushan.cn/blog/intentdna-route-reset-from-platform-to-workflow-cli) | [历史 Online 看板](https://online-plum-one.vercel.app)
 
-## 当前路线
+> **项目状态 / Project status:** 本项目后续不再进行独立扩展；新的研发自动化能力将以独立 CLI 等工具的形式在其他仓库开发。 / This project will not be expanded independently; new engineering automation capabilities will be developed in other repositories as standalone CLIs and related tools.
 
-IntentDNA 的目标没有变：让个人和团队完成工作的专业方法能够被复用、验证、分发和持续改进。改变的是交付顺序。
+IntentDNA 把一套可重复的 AI 工作方法描述为版本化的 `.dna.yaml` 资产，并提供确定性的校验、编译、同步、运行和证据能力。你可以用它定义步骤、角色、权限、交接、检查点、重试和验证规则，再把工作流同步到本地 Harness 中执行。
 
-过去我们优先建设跨 Harness 声明、编译器、Adapter、Hooks、资产市场、组织治理和 Online 控制面。工程证明了这些机制可以成立，但没有证明团队应先采用一套平台。现在的路线是：
+## 它解决什么问题
+
+AI 编程工作流经常散落在 Prompt、对话、个人经验和临时脚本中。IntentDNA 将这些方法变成可以审查和复用的工程资产：
+
+- 用 YAML 明确适用场景、输入、输出、角色、步骤和边界；
+- 在执行前确定性校验结构、引用、权限和工作流约束；
+- 同步为 Codex Skills、Claude agents/skills/hooks 等 Harness 资产；
+- 用 Controller 支持持久运行、并行步骤、检查点、审查、重试和恢复；
+- 打包、签名并验证本地运行证据；
+- 从真实 Skills 和源码提炼候选资产，同时保留来源与许可证信息。
+
+核心边界是：
 
 ```text
-真实研发任务
-  -> 专用工作流 CLI
-  -> 真实使用、测试与评估
-  -> 一批有效的工作流资产
-  -> 从已经出现的重复问题中提炼公共协议和治理能力
+.dna.yaml 资产 = 工作流定义
+dna validate / compile / sync = 确定性工具链
+Codex CLI / Claude Code / provider = 实际执行环境
+本地运行记录与 verifier = 完成证据
+Online = 历史看板与可选投影，不是运行时
 ```
 
-因此，IntentDNA 现在是一个 **workflow CLI foundation**，而不是通用 Agent 平台、统一工作流 DSL 或 SaaS 执行器。优先级是先结束一个具体任务，再考虑跨工具抽象。
-
-完整背景见《[IntentDNA 路线重置：从平台优先到工作流 CLI](https://wanglushan.cn/blog/intentdna-route-reset-from-platform-to-workflow-cli)》。此前的 [IntentDNA Online 看板](https://online-plum-one.vercel.app) 作为历史产品证据继续保留；它展示了资产选择、安装指引和证据投影，但不是当前运行时，也不执行工作流。
+IntentDNA 不重造通用模型循环，不把 Online 或 Supabase 当作本地权限来源，也不接受模型的“已完成”声明替代验证结果。
 
 ## 安装
 
@@ -29,74 +38,148 @@ npm install --global intentdna@latest
 dna --version
 ```
 
-Claude Code 用户还可以安装随 npm 同版本发布的插件：
+Claude Code 插件与 npm 包使用相同版本发布：
 
 ```bash
 claude plugin marketplace add lushan1314/intentdna-repo
 claude plugin install intentdna@intentdna
 ```
 
-## 快速开始
+## 使用现有工作流
 
-从一个真实任务和一个现有资产开始：
+先查看内置资产，再在目标项目中初始化和同步：
 
 ```bash
-dna quickstart --asset flutter-rewrite --sync-target codex
 dna templates list
 dna templates show flutter-rewrite
+dna quickstart --asset flutter-rewrite --sync-target codex
+
 dna init --template flutter-rewrite
 dna sync --codex --dry-run
 dna sync --codex
 ```
 
-在 Codex CLI 或 Claude Code 中运行同步后的工作流。执行后可在本地打包、签名并验证证据：
+`dna quickstart` 只输出指引；`dna sync --dry-run` 预览文件变更；`dna sync` 才会写入 Harness 资产，但不会自动运行工作流。
+
+同步完成后，在对应项目中使用 Codex CLI 或 Claude Code 调用生成的 Skill、agent 或工作流入口。检查同步状态：
 
 ```bash
-dna report package --json
-dna report sign --json
-dna report verify <envelope-path> --json
+dna sync status
+dna sync doctor
 ```
 
-`dna quickstart` 只打印指引；`dna sync` 只安装和同步资产。真正的执行、权限判断和验证仍发生在本地项目与所选 Harness 中。
+## 打造自己的工作流
 
-## 能力边界
+IntentDNA 将“理解人的意图”和“确定性验证”分开：外部 Agent 负责访谈并编写 YAML，`dna` 负责提供完整创作契约并验证结果，人负责确认是否保存和采用。
 
-IntentDNA 当前保留并继续验证这些能力：
+### 1. 生成创作包
 
-- 将任务边界、输入输出、工具、人工确认点和验证器封装为版本化工作流资产；
-- 把资产同步到 Codex CLI 和 Claude Code 等执行环境；
-- 用 Controller、检查点、审查和恢复机制完成可持续运行；
-- 记录本地运行事实，签名并验证证据；
-- 从真实 Skills、脚本和源码中提炼候选资产，并保留来源和许可证信息。
+```bash
+dna workflow authoring-packet "描述要沉淀的重复工作方法" \
+  --project-dir /path/to/project \
+  --json
+```
 
-IntentDNA 不做这些事情：
+创作包包含字段约束、红线、真实示例、访谈问题、校验循环和交接边界。这个命令不调用模型，也不生成或保存 YAML。
+
+### 2. 让 Agent 编写候选资产
+
+候选文件放在：
+
+```text
+.dna/authoring/<asset>.candidate.dna.yaml
+```
+
+一个工作流资产通常包括：
+
+- `id`、`name`、`version`、`purpose`、`maturity`；
+- `fit`、`not_for`、`inputs`、`outputs`；
+- `roles` 及其工具权限和读写范围；
+- `workflow` 或 `workflows` 中的步骤、依赖、交接、检查点和重试；
+- `boundaries`、`evidence` 和 `lineage`。
+
+### 3. 校验并修订
+
+```bash
+dna validate .dna/authoring/<asset>.candidate.dna.yaml
+```
 
-- 不重造通用模型循环或 Agent runtime；
-- 不要求所有工作流先进入统一 DSL；
-- 不把 Online、市场或 Supabase 当作本地运行权限的来源；
-- 不用模型的“已完成”声明替代确定性验证。
+Agent 只根据校验诊断和已确认的人类意图修改 YAML，直到校验通过。`dna` 不推断目标，也不会自动放宽边界。
 
-## Online 与历史资产
+### 4. 人工确认并加入项目资产库
 
-当前权威关系是：
+确认候选内容后，将同一份 YAML 保存为：
 
 ```text
-本地 workflow CLI / dna = 运行事实
-Harness = 实际执行环境
-Online = 历史看板与可选的资产/证据投影
-archive = 路线演进证据，不是当前产品承诺
+.dna/templates/<asset>.dna.yaml
+```
+
+然后检查资产并预览同步结果：
+
+```bash
+dna assets show <asset> --project-dir /path/to/project
+dna assets inspect <asset> --project-dir /path/to/project
+dna sync --project-dir /path/to/project --dry-run
+dna sync --project-dir /path/to/project --codex --dry-run
+```
+
+保存、采用和同步是彼此独立的人工决策。校验通过不代表资产已被采用或执行。
+
+## 运行与证据
+
+工作流可以在同步后的 Harness 中执行，也可以使用 IntentDNA Controller 的运行入口。可用命令以当前 CLI 帮助为准：
+
+```bash
+dna run --help
+dna run status <run-id>
+dna run inspect <run-id>
+dna run resume <run-id>
 ```
 
-仓库只保留两类文档：少量当前权威文档，以及带日期和索引的历史档案。会话状态、临时提示词、执行日志、IDE workspace 和 `.omx/.omc/.claude/.codex` 运行期文件不进入版本控制。
+真实执行后，可打包、签名并验证本地证据：
+
+```bash
+dna report package --json
+dna report sign --package .dna/reports/latest-report-package.json \
+  --output .dna/reports/ingest/local-envelope.json \
+  --json
+dna report verify .dna/reports/ingest/local-envelope.json --json
+```
+
+云端上传是可选能力，并且必须先通过凭证和部署检查。本地工作流的运行事实不会因为上传而转移到云端。
+
+## 内置工作流类型
+
+仓库包含多种可直接查看和复用的资产，包括：
+
+- Flutter 重写、行为锁定、诊断与修复审查；
+- 代码审查、TDD、QA、调试、迁移和发布证据；
+- 并行子 Agent、多角色交接和持久执行；
+- 研究改进、工作流创作和资产提炼；
+- 安全开发、事故响应和组织基线。
+
+使用 `dna templates list` 查看当前安装版本中的完整目录。
+
+## 路线转换与历史
+
+IntentDNA 曾优先建设跨 Harness 抽象、组织治理、市场和 Online 控制面。实践表明，这些机制具备工程可行性，但平台优先并不是后续研发自动化最合适的起点。
+
+完整反思见《[IntentDNA 路线重置：从平台优先到工作流 CLI](https://wanglushan.cn/blog/intentdna-route-reset-from-platform-to-workflow-cli)》。
+
+[IntentDNA Online 看板](https://online-plum-one.vercel.app) 作为历史产品证据继续保留，用于展示当时的资产选择、安装指引和证据投影。它不执行工作流。
+
+仓库中的当前文档和历史证据采用两态管理：少量 living docs 描述当前实现，带日期和索引的 `docs/archive/` 保存路线演进；本地会话状态、临时提示词、运行日志和 IDE workspace 不进入版本控制。
 
 ## 文档
 
 - [快速开始](docs/quickstart.md)
+- [工作流资产格式](docs/architecture/c1-asset-format.md)
+- [工作流创作包契约](docs/architecture/c3-authoring-packet.md)
+- [Controller 与 worker 运行时](docs/architecture/controller-worker-runtime.md)
 - [工作流运行时清单](docs/architecture/workflow-runtime-manifest.md)
+- [本地证据捕获](docs/architecture/c5-evidence-capture.md)
 - [产品重置 ADR](docs/decisions/ADR-001-product-reset.md)
-- [历史档案索引](docs/archive/2026H1/INDEX.md)
-- [2026 H2 路线分享](docs/archive/2026H2/sharing/2026-08-01-build-your-workflow-cli.md)
-- [v1.9.0 发布说明](docs/archive/2026H2/release-notes/v1.9.0.md)
+- [文档与历史档案索引](docs/README.md)
 
 ## License
 
@@ -108,27 +191,34 @@ Apache-2.0
 
 ## English
 
-**Turn real, repeated, verifiable AI work methods into independently executable workflow CLIs.**
+**Declarative workflow assets, synchronization, and runtime tooling for AI coding harnesses such as Codex CLI and Claude Code.**
 
 [Route reset article (Chinese)](https://wanglushan.cn/blog/intentdna-route-reset-from-platform-to-workflow-cli) | [Historical Online dashboard](https://online-plum-one.vercel.app)
 
-### Current Direction
+IntentDNA describes repeatable AI work methods as versioned `.dna.yaml` assets and provides deterministic validation, compilation, synchronization, runtime, and evidence tooling. A workflow can define steps, roles, permissions, handoffs, checkpoints, retries, and verifiers before being synchronized into a local harness.
+
+### What It Solves
+
+AI coding workflows often remain scattered across prompts, conversations, personal experience, and temporary scripts. IntentDNA turns those methods into reviewable engineering assets:
 
-IntentDNA still exists to make professional work methods reusable, testable, distributable, and improvable. What changed is the delivery order.
+- define fit, inputs, outputs, roles, steps, and boundaries in YAML;
+- validate structure, references, permissions, and workflow constraints deterministically;
+- synchronize Codex Skills, Claude agents/skills/hooks, and other harness assets;
+- support durable runs, parallel steps, checkpoints, reviews, retries, and recovery;
+- package, sign, and verify local runtime evidence;
+- refine candidates from real skills and source while retaining provenance and license data.
 
-The project previously led with cross-harness declarations, a compiler, adapters, hooks, a marketplace, organization governance, and an Online control plane. That work proved the mechanisms could be built, but it did not prove that teams should adopt a platform first. The current route is:
+The authority model is:
 
 ```text
-real engineering task
-  -> task-specific workflow CLI
-  -> real use, tests, and evaluation
-  -> a portfolio of useful workflow assets
-  -> shared protocols and governance extracted from observed repetition
+.dna.yaml asset = workflow definition
+dna validate / compile / sync = deterministic toolchain
+Codex CLI / Claude Code / provider = execution environment
+local run records and verifiers = completion evidence
+Online = historical dashboard and optional projection, not runtime
 ```
 
-IntentDNA is therefore a **workflow CLI foundation**, not a generic agent platform, universal workflow DSL, or SaaS executor. Finish one concrete task first; abstract across tools only when real repetition justifies it.
-
-The full rationale is documented in [IntentDNA Route Reset: From Platform to Workflow CLI](https://wanglushan.cn/blog/intentdna-route-reset-from-platform-to-workflow-cli). The previous [IntentDNA Online dashboard](https://online-plum-one.vercel.app) remains available as historical product evidence. It demonstrates asset selection, setup guidance, and evidence projection, but it is not the current runtime and does not execute workflows.
+IntentDNA does not rebuild the generic model loop, make Online or Supabase the authority for local execution, or accept a model's completion claim as a substitute for verification.
 
 ### Install
 
@@ -137,74 +227,124 @@ npm install --global intentdna@latest
 dna --version
 ```
 
-Claude Code users can install the plugin released at the same version as npm:
+The Claude Code plugin is released at the same version as the npm package:
 
 ```bash
 claude plugin marketplace add lushan1314/intentdna-repo
 claude plugin install intentdna@intentdna
 ```
 
-### Quick Start
+### Use An Existing Workflow
 
-Start with a real task and an existing asset:
+Inspect built-in assets, then initialize and synchronize one inside the target project:
 
 ```bash
-dna quickstart --asset flutter-rewrite --sync-target codex
 dna templates list
 dna templates show flutter-rewrite
+dna quickstart --asset flutter-rewrite --sync-target codex
+
 dna init --template flutter-rewrite
 dna sync --codex --dry-run
 dna sync --codex
 ```
 
-Run the synced workflow in Codex CLI or Claude Code. Then package, sign, and verify local evidence:
+`dna quickstart` only prints guidance. `dna sync --dry-run` previews file changes. `dna sync` writes harness assets but does not execute the workflow.
+
+Use the generated Skill, agent, or workflow entry point from Codex CLI or Claude Code. Inspect synchronization state with:
 
 ```bash
-dna report package --json
-dna report sign --json
-dna report verify <envelope-path> --json
+dna sync status
+dna sync doctor
 ```
 
-`dna quickstart` only prints guidance. `dna sync` only installs and synchronizes assets. Execution, permission decisions, and verification remain in the local project and selected harness.
+### Author A Workflow
+
+IntentDNA separates probabilistic intent interpretation from deterministic validation. An external agent interviews the human and authors YAML; `dna` emits the complete authoring contract and validates the result; the human decides whether the candidate is saved and adopted.
+
+Generate an authoring packet:
 
-### Boundaries
+```bash
+dna workflow authoring-packet "describe the repeated work method" \
+  --project-dir /path/to/project \
+  --json
+```
 
-IntentDNA continues to validate these capabilities:
+The packet includes field constraints, red lines, real examples, interview questions, the validation loop, and handoff boundaries. It does not call a model or create YAML.
 
-- package task boundaries, inputs, outputs, tools, human gates, and verifiers as versioned workflow assets;
-- synchronize assets into execution environments such as Codex CLI and Claude Code;
-- support durable runs with controllers, checkpoints, reviews, and recovery;
-- capture local run facts and produce signed, verifiable evidence;
-- refine candidate assets from real skills, scripts, and source while retaining provenance and license data.
+Have the external agent write the candidate to:
 
-IntentDNA does not:
+```text
+.dna/authoring/<asset>.candidate.dna.yaml
+```
 
-- rebuild the generic model loop or agent runtime;
-- require every workflow to enter a universal DSL first;
-- make Online, a marketplace, or Supabase the authority for local execution;
-- accept a model's completion claim as a substitute for deterministic verification.
+A workflow asset normally describes metadata, fit, inputs and outputs, roles and permissions, steps and dependencies, handoffs, checkpoints, retries, boundaries, evidence, and lineage.
 
-### Online And History
+Validate and revise it:
 
-The current authority model is:
+```bash
+dna validate .dna/authoring/<asset>.candidate.dna.yaml
+```
 
-```text
-local workflow CLI / dna = runtime truth
-harness = execution environment
-Online = historical dashboard and optional asset/evidence projection
-archive = route-evolution evidence, not current product commitment
+After explicit human confirmation, save the same YAML under `.dna/templates/<asset>.dna.yaml`, then inspect it and preview synchronization:
+
+```bash
+dna assets show <asset> --project-dir /path/to/project
+dna assets inspect <asset> --project-dir /path/to/project
+dna sync --project-dir /path/to/project --dry-run
+dna sync --project-dir /path/to/project --codex --dry-run
 ```
 
-The repository keeps only a small living documentation set and dated, indexed historical archives. Session state, temporary prompts, execution logs, IDE workspace data, and runtime files under `.omx/.omc/.claude/.codex` do not belong in version control.
+Saving, adopting, and synchronizing are separate human decisions. Successful validation does not mean the asset has been adopted or executed.
+
+### Runtime And Evidence
+
+Workflows can execute in the synchronized harness or through the IntentDNA Controller. Use the installed CLI help as the command authority:
+
+```bash
+dna run --help
+dna run status <run-id>
+dna run inspect <run-id>
+dna run resume <run-id>
+```
+
+After a real run, package, sign, and verify local evidence:
+
+```bash
+dna report package --json
+dna report sign --package .dna/reports/latest-report-package.json \
+  --output .dna/reports/ingest/local-envelope.json \
+  --json
+dna report verify .dna/reports/ingest/local-envelope.json --json
+```
+
+Cloud upload is optional and credential-gated. Uploading evidence does not move runtime authority away from the local workflow.
+
+### Included Workflow Categories
+
+The repository includes assets for Flutter rewrites and repair reviews, code review, TDD, QA, debugging, migrations, release evidence, parallel subagents, multi-role handoffs, durable execution, research improvement, workflow authoring, asset refinement, secure development, incident response, and organization baselines.
+
+Run `dna templates list` for the complete catalog in the installed version.
+
+### Route Reset And History
+
+IntentDNA previously led with cross-harness abstraction, organization governance, a marketplace, and an Online control plane. Those experiments demonstrated engineering feasibility, but platform-first development is not the chosen starting point for the next phase of task-specific automation.
+
+The rationale is documented in [IntentDNA Route Reset: From Platform to Workflow CLI](https://wanglushan.cn/blog/intentdna-route-reset-from-platform-to-workflow-cli).
+
+The [IntentDNA Online dashboard](https://online-plum-one.vercel.app) remains available as historical product evidence. It demonstrates the earlier asset selection, setup guidance, and evidence projection work; it does not execute workflows.
+
+Current documentation and historical evidence follow a two-state model: a small living set describes the implementation, while dated, indexed files under `docs/archive/` preserve route history. Local session state, temporary prompts, runtime logs, and IDE workspace data are excluded from version control.
 
 ### Documentation
 
 - [Quickstart](docs/quickstart.md)
+- [Workflow asset format](docs/architecture/c1-asset-format.md)
+- [Workflow authoring packet](docs/architecture/c3-authoring-packet.md)
+- [Controller and worker runtime](docs/architecture/controller-worker-runtime.md)
 - [Workflow runtime manifest](docs/architecture/workflow-runtime-manifest.md)
+- [Local evidence capture](docs/architecture/c5-evidence-capture.md)
 - [Product reset ADR](docs/decisions/ADR-001-product-reset.md)
-- [Historical archive index](docs/archive/2026H1/INDEX.md)
-- [2026 H2 route-sharing note](docs/archive/2026H2/sharing/2026-08-01-build-your-workflow-cli.md)
-- [v1.9.0 release notes](docs/archive/2026H2/release-notes/v1.9.0.md)
+- [Documentation and archive index](docs/README.md)
 
 ### License
 

package/dist/cli/commands/run-lifecycle.d.ts

@@ -1,30 +1,42 @@
-import type { JsonValue, RunId, TaskState } from "../../runtime/run-contracts.js";
-import type { RunController } from "../../runtime/run-controller.js";
-import type { DurableRunStore, RunRecord, RunStatus, RunStoreSnapshot } from "../../runtime/run-store.js";
-export type RunLifecycleAction = "start" | "status" | "inspect" | "resume" | "cancel";
+import type { CanonicalActivationState, JsonValue, RunId, TaskState } from "../../runtime/run-contracts.js";
+import type { RunAuditView, RunView } from "../../runtime/canonical-run-service.js";
+import { type DurableRunStore, type RunRecord, type RunStatus, type RunStoreSnapshot } from "../../runtime/run-store.js";
+export type RunLifecycleAction = "start" | "status" | "inspect" | "resume" | "cancel" | "close";
 interface LifecycleOutputOptions {
     readonly json?: boolean;
 }
-export interface RunLifecycleStartOptions extends LifecycleOutputOptions {
-    readonly controller: RunController;
-    readonly run: RunRecord;
-}
 export interface RunLifecycleStatusOptions extends LifecycleOutputOptions {
     readonly run_store: DurableRunStore;
     readonly run_id: RunId;
+    readonly service?: CanonicalLifecycleService;
+    readonly service_factory?: () => Promise<CanonicalLifecycleService>;
 }
 export interface RunLifecycleInspectOptions extends LifecycleOutputOptions {
     readonly run_store: DurableRunStore;
     readonly run_id: RunId;
+    readonly service?: CanonicalLifecycleService;
+    readonly service_factory?: () => Promise<CanonicalLifecycleService>;
 }
 export interface RunLifecycleResumeOptions extends LifecycleOutputOptions {
-    readonly controller: RunController;
+    readonly run_store: DurableRunStore;
     readonly run_id: RunId;
+    readonly service?: CanonicalLifecycleService;
+    readonly service_factory?: () => Promise<CanonicalLifecycleService>;
+    readonly drive?: (service: CanonicalLifecycleService, run_id: RunId) => Promise<RunView>;
 }
 export interface RunLifecycleCancelOptions extends LifecycleOutputOptions {
-    readonly controller: RunController;
+    readonly run_store: DurableRunStore;
     readonly run_id: RunId;
     readonly reason?: string | null;
+    readonly service?: CanonicalLifecycleService;
+    readonly service_factory?: () => Promise<CanonicalLifecycleService>;
+}
+export type RunLifecycleCloseOptions = RunLifecycleCancelOptions;
+export interface CanonicalLifecycleService {
+    inspect(run_id: RunId): Promise<RunView>;
+    inspectAudit(run_id: RunId): Promise<RunAuditView>;
+    resume(run_id: RunId): Promise<RunView>;
+    cancel(run_id: RunId, reason?: string | null): Promise<RunView>;
 }
 export interface RunStatusProjection {
     readonly schema_version: "intentdna.run_status.v1";
@@ -41,6 +53,24 @@ export interface RunStatusProjection {
     readonly created_at: string;
     readonly updated_at: string;
 }
+export interface CanonicalRunStatusProjection {
+    readonly schema_version: "intentdna.run_status.v2";
+    readonly ledger_schema_version: 2;
+    readonly run_id: RunId;
+    readonly target: RunView["run"]["target"];
+    readonly plan_id: string;
+    readonly binding_id: string;
+    readonly status: RunStatus;
+    readonly record_version: number;
+    readonly activation_counts: Readonly<Record<CanonicalActivationState, number>>;
+    readonly attempt_count: number;
+    readonly result_count: number;
+    readonly cancellation: RunView["run"]["cancellation"];
+    readonly terminal: RunView["run"]["terminal"];
+    readonly created_at: string;
+    readonly updated_at: string;
+}
+export type AnyRunStatusProjection = RunStatusProjection | CanonicalRunStatusProjection;
 export interface DurableHandoffInspection {
     readonly event_id: string;
     readonly sequence: number;
@@ -65,9 +95,29 @@ export interface RunInspectionProjection {
     readonly ledger_created_at: string;
     readonly ledger_updated_at: string;
 }
-export declare function runLifecycleStart(options: RunLifecycleStartOptions): Promise<number>;
+export interface CanonicalRunInspectionProjection {
+    readonly schema_version: "intentdna.run_inspection.v2";
+    readonly ledger_schema_version: 2;
+    readonly record_version: number;
+    readonly run: RunAuditView["run"];
+    readonly activations: RunAuditView["activations"];
+    readonly attempts: RunAuditView["attempts"];
+    readonly claims: RunAuditView["claims"];
+    readonly events: RunAuditView["events"];
+    readonly result_refs: RunAuditView["results"];
+    readonly ledger_created_at: string;
+    readonly ledger_updated_at: string;
+}
+interface LifecycleOperationProjection {
+    readonly schema_version: "intentdna.run_lifecycle_operation.v1";
+    readonly action: Exclude<RunLifecycleAction, "status" | "inspect">;
+    readonly status: AnyRunStatusProjection;
+}
+export declare function canonicalViewStatusProjection(view: RunView): CanonicalRunStatusProjection;
+export declare function writeCanonicalOperation(action: LifecycleOperationProjection["action"], view: RunView, json: boolean): void;
 export declare function runLifecycleStatus(options: RunLifecycleStatusOptions): Promise<number>;
 export declare function runLifecycleInspect(options: RunLifecycleInspectOptions): Promise<number>;
 export declare function runLifecycleResume(options: RunLifecycleResumeOptions): Promise<number>;
 export declare function runLifecycleCancel(options: RunLifecycleCancelOptions): Promise<number>;
+export declare function runLifecycleClose(options: RunLifecycleCloseOptions): Promise<number>;
 export {};