@dawipong/opcflow 0.2.0 → 0.3.0

package/README.en.md

@@ -33,7 +33,7 @@ contracts, thumbs-up/down outputs, and answer rulings**.
 - **Change propagation** — `sync` reconciles → invalidate → dispatch review along the graph (deduped)
 - **QA loop** — fail → auto rework → auto re-verify, no human in the loop
 - **Write-gate hooks** — block agents from editing an approved contract (observe mode by default)
-- **Feedback evolution** — 👍👎 and QA verdicts, half-life weighted → skill candidates / Red Flags
+- **Feedback evolution** — 👍👎 and QA verdicts, half-life weighted → candidates (AI decides skill / rule / memory) / Red Flags
 - **Multi-platform** — one definition generates each platform's agent + MCP + hooks (see [PLATFORMS.md](PLATFORMS.md))
 - **Visual opcflow** — tree + markdown/mermaid/prototype/code rendering + review-queue diff + live SSE
 
@@ -55,7 +55,7 @@ complementary:
 | Roles | Essentially single-flow (one agent implements) | Five-role pipeline, each with its own gate and output channel |
 | Construction constraints | The spec is advisory | Gates block tasks whose upstream isn't approved; write-gate blocks edits to approved contracts; protocol lints as checkpoints |
 | Acceptance | Not covered | Two-stage QA + fail→rework→re-verify auto loop |
-| Evolution | Not covered | 👍👎/verdict half-life weighting → skill candidates / Red Flags |
+| Evolution | Not covered | 👍👎/verdict half-life weighting → candidates (skill/rule/memory) / Red Flags |
 
 In one line: **Spec Kit treats the spec as "context for the agent"; opcflow treats it as "an
 approved contract that can go stale, propagate, and block construction," and governs the entire drift
@@ -141,94 +141,30 @@ never silently deviate.
 > `docs/design/prototypes/<endpoint>/<module>/<page>.html` — once `scan` registers it, it goes through
 > the usual 👍 release. Both paths are equivalent in the system; both must pass a user 👍.
 
-## CLI Commands & Parameters
+## CLI Commands
 
-Every command: `opcflow <command> [args]` (after global install; or `npx -y @dawipong/opcflow <command>` without installing). Global `--project=<path>` sets the project
-root (otherwise it searches upward for `workbench.config.json`). File-path arguments are separated with
-`--` (e.g. `submit --actor=x -- <path>`).
+`opcflow <command> [args]` (after a global install; or `npx -y @dawipong/opcflow <command>` without installing). Approval actions (approve/reject) are human-only; the AI uses the MCP `wb_*` typed tools.
 
-**Tasks**
+**Each command's purpose, when to use it, and parameters are in [COMMANDS.en.md](COMMANDS.en.md).** At a glance:
 
-| Command | Purpose | Parameters |
-| --- | --- | --- |
-| `list` | list tasks | `--status` `--assignee` `--module` `--role` `--endpoint` `--type` `--stale=true` |
-| `show <id>` | task detail (events / outputs / stale) | `--json=true` |
-| `create` | create a task | required `--role` `--creator`; opt `--module` `--endpoint` `--page` `--type` `--assignee` `--content` |
-| `claim <id>` | claim (passes the gate, auto-snapshots deps) | `--assignee=<role>` |
-| `update <id>` | change status | `--status` `--operator`; `--force=true` bypasses the stale block (traced) |
-| `remove <id>` | delete a task | `--operator`; `--force=true` |
-| `record <id> "note"` | add a note | `--operator` |
-| `input <id> -- <path>` | declare a dependency read outside the gate | `--operator` |
-
-**Outputs**
-
-| Command | Purpose | Parameters |
-| --- | --- | --- |
-| `output -- <path>` | register an output (auto-links your claimed task) | required `--role` `--endpoint`; opt `--module` `--page` `--task` |
-| `artifacts` | list artifacts | `--module` `--endpoint` `--page` `--kind` |
-| `scan` | full scan & register (docs + codeRoots), derive DAG edges | `--actor` |
-| `move --from=<> --to=<>` | move an artifact (keeps id, keeps approval) | `--actor` |
-
-**Trust**
-
-| Command | Purpose | Parameters |
-| --- | --- | --- |
-| `submit -- <path>` | submit for review | `--actor` |
-| `approve -- <path>` | approve | `--actor`; `--trivial=true` for tweaks: re-bless downstream + close derived reviews |
-| `reject -- <path>` | reject | `--actor` `--reason` |
-| `feedback -- <path>` | feedback / 👍👎 (release prototype) | `--actor` `--verdict=+1\|-1`; opt `--comment` `--task` |
-| `dispute -- <path>` | trace an objection to approved content and stop for a ruling | `--actor` `--reason` |
-| `queue` | review queue (pending + invalidated) | — |
-| `sync` | reconcile: invalidation propagation + dispatch review | `--actor` |
-
-**Flow**
-
-| Command | Purpose | Parameters |
-| --- | --- | --- |
-| `plan` | dispatch downstream tasks once contracts are approved | `--module`; opt `--creator` |
-| `qa <id>` | record acceptance result (fail auto-dispatches rework) | `--result=pass\|fail` `--operator`; fail needs `--reason` |
-| `audit` | module contract reconciliation report (settlement status) | `--module` |
-| `graph` | emit a Mermaid dependency graph | `--module` |
-| `lint` | run protocol lints | opt `--role` `--endpoint` |
-| `events` | event stream | `--taskId`/`<id>` `--module` `--event` `--limit` `--json` |
-| `intake` | pull GitHub issues into the queue (standard lane / hotfix) | — |
-
-**Evolution / Maintenance**
-
-| Command | Purpose | Parameters |
-| --- | --- | --- |
-| `retro` | retrospective: skill candidates / Red Flags / approval throughput | opt `--module` `--json` |
-| `export` | export events/feedback as jsonl | — |
-| `init` | bootstrap a new project | `--endpoints` (required); `--platforms` `--model` `--writehooks=false` `--preset=false` `--hooks=false` |
-| `gen-agents` | regenerate agent definitions from templates | — |
-| `register-meta` | register meta artifacts (agent-def/skill/plan) | `--actor` |
-| `install-hooks` | install git hooks | — |
-| `migrate` | migrate a legacy DB | `--from=<path>` |
-
-> On the AI side, prefer the MCP `wb_*` typed tools (same source & transaction as the CLI). **approve/reject are deliberately not exposed to the AI — those are your actions.**
+- **Tasks** `list` · `show` · `create` · `claim` · `update` · `remove` · `record` · `input`
+- **Outputs** `output` · `artifacts` · `scan` · `move`
+- **Trust** `submit` · `approve` · `reject` · `feedback` · `dispute` · `queue` · `sync`
+- **Flow** `plan` · `qa` · `audit` · `graph` · `lint` · `events` · `intake`
+- **Evolution / Maintenance** `retro` · `export` · `init` · `gen-agents` · `register-meta` · `install-hooks` · `migrate`
+- **Service & Integration** `serve` · `mcp` · `hook` · `postcommit` (mostly auto-invoked by platform / git)
 
 ## Configuration (workbench.config.json)
 
+Generated by `init`, hand-edited afterwards — each project's coordinate system and discipline switches. **Every field's purpose, default, and when to tune it is in [CONFIG.en.md](CONFIG.en.md).**
+
 ```jsonc
 {
-  "platforms": ["claude", "cursor"],          // target platforms (picked by init), default ["claude"]
+  "platforms": ["claude", "cursor"],          // target platforms
   "endpoints": ["service", "web"],            // your endpoints
-  "pipeline": ["product-manager", "architect", "designer", "developer", "qa"], // role pipeline (excluded roles get no tasks)
-  "codeRoots": {                              // [required] code dir per endpoint, {module} placeholder
-    "service": ["service/src/modules/{module}"],
-    "web": ["web/src/pages/{module}"]
-  },
-  "moduleMapping": { "userProfile": "user" }, // merge fine modules into coarse ones
-  "machineChecks": { "enabled": false,        // run on developer complete
-    "service": ["cd service && npx tsc --noEmit"] },
-  "protocolLints": [                          // machine-checkable conventions → lints (violations block complete)
-    { "name": "no-page-size", "grep": "pageSize", "paths": ["service/src"] }
-  ],
-  "gates": {
-    "approvalMode": "warn",                   // warn = unreviewed only warns / enforce = blocks
-    "writeGate": "observe"                     // off / observe (log only) / enforce (block edits to approved contracts)
-  },
-  "git": { "taskTrailer": "off", "trailerKey": "Task" }  // on = inject a Task:#id attribution trailer on commit
+  "pipeline": ["product-manager", "architect", "designer", "developer", "qa"],
+  "codeRoots": { "service": ["service/src/modules/{module}"] },  // [required] code dir per endpoint, {module} placeholder
+  "gates": { "approvalMode": "warn", "writeGate": "observe" }    // approval / write-gate discipline levels
 }
 ```
 

package/README.md

@@ -31,7 +31,7 @@
 - **变更传播** —— `sync` 对账 → 失效 → 沿图派 review(去重)
 - **QA 闭环** —— fail → 自动 rework → 自动复验,不消耗人
 - **写门禁 hooks** —— agent 改到已 approved 契约时拦截(默认 observe 观察期)
-- **反馈进化** —— 👍👎 与 QA verdict 半衰期加权 → skill 候选 / Red Flags
+- **反馈进化** —— 👍👎 与 QA verdict 半衰期加权 → 经验候选(由 AI 判断沉淀为 skill / 规则 / 记忆)/ Red Flags
 - **多平台** —— 一套定义生成四家平台各自的 agent + MCP + hooks(见 [PLATFORMS.md](PLATFORMS.md))
 - **可视化工作台** —— 树 + markdown/mermaid/原型/代码渲染 + 待审队列 diff + SSE 实时
 
@@ -51,7 +51,7 @@ opcflow 接管 **spec 之后的执行与验证**,两者是不同层次、可互
 | 角色 | 基本单流程(一个 agent 实现) | 五角色流水线,各有 gate 与产出通道 |
 | 施工约束 | spec 是建议 | gate 拦住上游未批的任务;写门禁拦改已批契约;协议 lint 卡点 |
 | 验收 | 不涉及 | 两段式 QA + fail→rework→复验自动闭环 |
-| 进化 | 不涉及 | 👍👎/verdict 半衰期加权 → skill 候选 / Red Flags |
+| 进化 | 不涉及 | 👍👎/verdict 半衰期加权 → 经验候选(skill/规则/记忆)/ Red Flags |
 
 一句话:**Spec Kit 把 spec 当"给 agent 的上下文";opcflow 把 spec 当"会失效、会传播、卡得住施工
 的已批契约",并管住从契约到代码到验收的整条漂移。**
@@ -129,93 +129,30 @@ flowchart LR
 > `docs/design/prototypes/<端>/<模块>/<页面>.html`,`scan` 登记后照常走 👍 放行——两条路径产出的原型
 > 在体系里等价,都要过用户 👍。
 
-## CLI 命令参数
+## CLI 命令
 
-所有命令:`opcflow <命令> [参数]`(全局装后即用;或 `npx -y @dawipong/opcflow <命令>` 免装)。全局 `--project=<路径>` 指定项目根(缺省向上找
-`workbench.config.json`)。文件路径参数用 `--` 分隔(如 `submit --actor=x -- <路径>`)。
+`opcflow <命令> [参数]`(全局装后即用;或 `npx -y @dawipong/opcflow <命令>` 免装)。审批类动作(approve/reject)只给人,AI 侧走 MCP 的 `wb_*` typed tools。
 
-**任务**
+**每条命令的作用、使用场景与参数见 [COMMANDS.md](COMMANDS.md)。** 速览:
 
-| 命令 | 作用 | 参数 |
-| --- | --- | --- |
-| `list` | 列任务 | `--status` `--assignee` `--module` `--role` `--endpoint` `--type` `--stale=true` |
-| `show <id>` | 任务详情(含事件/产出/stale) | `--json=true` |
-| `create` | 建任务 | 必填 `--role` `--creator`;选 `--module` `--endpoint` `--page` `--type` `--assignee` `--content` |
-| `claim <id>` | 领取(过 gate,自动快照依赖) | `--assignee=<角色>` |
-| `update <id>` | 改状态 | `--status` `--operator`;`--force=true` 越过 stale 拦截(留痕) |
-| `remove <id>` | 删任务 | `--operator`;`--force=true` |
-| `record <id> "备注"` | 加备注 | `--operator` |
-| `input <id> -- <路径>` | 补充申报 gate 之外读过的依赖 | `--operator` |
-
-**产出**
-
-| 命令 | 作用 | 参数 |
-| --- | --- | --- |
-| `output -- <路径>` | 登记产物(自动关联你领取的任务) | 必填 `--role` `--endpoint`;选 `--module` `--page` `--task` |
-| `artifacts` | 列产物 | `--module` `--endpoint` `--page` `--kind` |
-| `scan` | 全量扫描登记(docs + codeRoots),推导 DAG 边 | `--actor` |
-| `move --from=<> --to=<>` | 移动产物(保 id、不断审批) | `--actor` |
-
-**信任**
-
-| 命令 | 作用 | 参数 |
-| --- | --- | --- |
-| `submit -- <路径>` | 送审 | `--actor` |
-| `approve -- <路径>` | 批准 | `--actor`;`--trivial=true` 微调:re-bless 下游 + 关派生 review |
-| `reject -- <路径>` | 打回 | `--actor` `--reason` |
-| `feedback -- <路径>` | 反馈 / 👍👎(原型放行) | `--actor` `--verdict=+1\|-1`;选 `--comment` `--task` |
-| `dispute -- <路径>` | 对 approved 内容留痕异议,停下等裁决 | `--actor` `--reason` |
-| `queue` | 待审队列(pending + invalidated) | — |
-| `sync` | 对账:失效传播 + 沿图派 review | `--actor` |
-
-**流程**
-
-| 命令 | 作用 | 参数 |
-| --- | --- | --- |
-| `plan` | 契约全批后一键派发下游任务 | `--module`;选 `--creator` |
-| `qa <id>` | 记验收结果(fail 自动派 rework) | `--result=pass\|fail` `--operator`;fail 需 `--reason` |
-| `audit` | 模块契约对账报告(清算状态) | `--module` |
-| `graph` | 输出 Mermaid 依赖图 | `--module` |
-| `lint` | 跑协议 lint | 选 `--role` `--endpoint` |
-| `events` | 事件流 | `--taskId`/`<id>` `--module` `--event` `--limit` `--json` |
-| `intake` | 拉 GitHub issue 入队(标准道 / hotfix) | — |
-
-**进化 / 维护**
-
-| 命令 | 作用 | 参数 |
-| --- | --- | --- |
-| `retro` | 复盘:skill 候选 / Red Flag / 审批吞吐 | 选 `--module` `--json` |
-| `export` | 导出 events/feedback 为 jsonl | — |
-| `init` | 新项目引导 | `--endpoints`(必填);`--platforms` `--model` `--writehooks=false` `--preset=false` `--hooks=false` |
-| `gen-agents` | 从模板重生成 agent 定义 | — |
-| `register-meta` | 注册元产物(agent-def/skill/plan) | `--actor` |
-| `install-hooks` | 安装 git hooks | — |
-| `migrate` | 迁移旧库 | `--from=<路径>` |
-
-> AI 侧优先用 MCP 的 `wb_*` typed tools(与 CLI 同源同事务);**审批 approve/reject 刻意不暴露给 AI——那是你的动作**。
+- **任务** `list` · `show` · `create` · `claim` · `update` · `remove` · `record` · `input`
+- **产出** `output` · `artifacts` · `scan` · `move`
+- **信任** `submit` · `approve` · `reject` · `feedback` · `dispute` · `queue` · `sync`
+- **流程** `plan` · `qa` · `audit` · `graph` · `lint` · `events` · `intake`
+- **进化 / 维护** `retro` · `export` · `init` · `gen-agents` · `register-meta` · `install-hooks` · `migrate`
+- **服务与集成** `serve` · `mcp` · `hook` · `postcommit`(多由平台 / git 自动调用)
 
 ## 配置(workbench.config.json)
 
+`init` 生成、之后手工编辑,是每个项目的坐标系与纪律开关。**每个字段的作用、默认值与何时调,见 [CONFIG.md](CONFIG.md)。**
+
 ```jsonc
 {
-  "platforms": ["claude", "cursor"],          // 目标平台(init 选),缺省 ["claude"]
+  "platforms": ["claude", "cursor"],          // 目标平台
   "endpoints": ["service", "web"],            // 你的端
-  "pipeline": ["product-manager", "architect", "designer", "developer", "qa"], // 角色流水线(不含的角色不派任务)
-  "codeRoots": {                              // 【必填】每端代码目录,{module} 占位
-    "service": ["service/src/modules/{module}"],
-    "web": ["web/src/pages/{module}"]
-  },
-  "moduleMapping": { "userProfile": "user" }, // 细模块归并到粗模块
-  "machineChecks": { "enabled": false,        // developer complete 时跑
-    "service": ["cd service && npx tsc --noEmit"] },
-  "protocolLints": [                          // 能机器查的约定 → 降级为 lint(违例阻断 complete)
-    { "name": "no-page-size", "grep": "pageSize", "paths": ["service/src"] }
-  ],
-  "gates": {
-    "approvalMode": "warn",                   // warn=未批只警告 / enforce=阻断
-    "writeGate": "observe"                     // off / observe(只记录)/ enforce(拦改 approved 契约)
-  },
-  "git": { "taskTrailer": "off", "trailerKey": "Task" }  // on=提交注入 Task:#id 归因 trailer
+  "pipeline": ["product-manager", "architect", "designer", "developer", "qa"],
+  "codeRoots": { "service": ["service/src/modules/{module}"] },  // 【必填】每端代码目录,{module} 占位
+  "gates": { "approvalMode": "warn", "writeGate": "observe" }    // 审批 / 写门禁纪律档位
 }
 ```
 

package/dist/cli.mjs

@@ -2528,7 +2528,7 @@ function deriveEdges(ctx) {
   return created;
 }
 function scanArtifacts(ctx, actor = "system") {
-  const summary = { registered: 0, refreshed: 0, edges: 0, skipped: [] };
+  const summary = { registered: 0, refreshed: 0, remapped: 0, edges: 0, skipped: [] };
   const files = [];
   for (const dir of Object.values(ctx.config.docs)) {
     const abs = join10(ctx.root, dir);
@@ -2537,7 +2537,12 @@ function scanArtifacts(ctx, actor = "system") {
   for (const p of ["ARCHITECTURE.md", "TECH.md"]) {
     if (existsSync9(join10(ctx.root, p))) files.push(p);
   }
-  const exists = ctx.db.prepare("SELECT id, content_hash FROM artifacts WHERE path = ?");
+  const exists = ctx.db.prepare(
+    "SELECT id, kind, module, endpoint, page, content_hash FROM artifacts WHERE path = ?"
+  );
+  const remap = ctx.db.prepare(
+    "UPDATE artifacts SET kind = ?, module = ?, endpoint = ?, page = ?, updated_at = CURRENT_TIMESTAMP WHERE id = ?"
+  );
   const insert = ctx.db.prepare(
     "INSERT INTO artifacts (kind, module, endpoint, page, path, content_hash) VALUES (?, ?, ?, ?, ?, ?)"
   );
@@ -2547,6 +2552,27 @@ function scanArtifacts(ctx, actor = "system") {
       const before = row.content_hash;
       const after = refreshArtifact(ctx, { id: row.id }, actor);
       if (after.content_hash !== before) summary.refreshed++;
+      if (row.kind !== kind || row.module !== coords.module || row.endpoint !== coords.endpoint || row.page !== coords.page) {
+        const tx2 = ctx.db.transaction(() => {
+          remap.run(kind, coords.module, coords.endpoint, coords.page, row.id);
+          logEvent(ctx.db, {
+            entityType: "artifact",
+            entityId: row.id,
+            event: "coords_remapped",
+            actor,
+            payload: {
+              path: relPath,
+              from: { kind: row.kind, module: row.module, endpoint: row.endpoint, page: row.page },
+              to: { kind, module: coords.module, endpoint: coords.endpoint, page: coords.page }
+            },
+            module: coords.module,
+            endpoint: coords.endpoint,
+            page: coords.page
+          });
+        });
+        tx2();
+        summary.remapped++;
+      }
       return;
     }
     const hash = hashPath(join10(ctx.root, relPath));
@@ -2655,8 +2681,8 @@ var init_qa_command = __esm({
 var retro_command_exports = {};
 __export(retro_command_exports, {
   AUTO_ACTOR_WEIGHT: () => AUTO_ACTOR_WEIGHT,
+  CANDIDATE_THRESHOLD: () => CANDIDATE_THRESHOLD,
   RED_FLAG_THRESHOLD: () => RED_FLAG_THRESHOLD,
-  SKILL_CANDIDATE_THRESHOLD: () => SKILL_CANDIDATE_THRESHOLD,
   approvalStats: () => approvalStats,
   distillFeedback: () => distillFeedback,
   exportEventLog: () => exportEventLog,
@@ -2670,6 +2696,8 @@ function parseDbTime(s) {
 function distillFeedback(ctx, opts = {}) {
   const now = (opts.now ?? /* @__PURE__ */ new Date()).getTime();
   const halfLifeMs = ctx.config.feedbackHalfLifeDays * 24 * 60 * 60 * 1e3;
+  const candidateThreshold = ctx.config.candidateThreshold ?? CANDIDATE_THRESHOLD;
+  const redFlagThreshold = ctx.config.redFlagThreshold ?? RED_FLAG_THRESHOLD;
   let query = `
     SELECT f.artifact_id, f.verdict, f.comment, f.actor, f.created_at,
            a.kind, a.module, a.endpoint, a.path
@@ -2710,13 +2738,13 @@ function distillFeedback(ctx, opts = {}) {
   for (const g of result) {
     g.posScore = Math.round(g.posScore * 100) / 100;
     g.negScore = Math.round(g.negScore * 100) / 100;
-    const pos = g.posScore >= SKILL_CANDIDATE_THRESHOLD;
-    const neg = g.negScore >= RED_FLAG_THRESHOLD;
+    const pos = g.posScore >= candidateThreshold;
+    const neg = g.negScore >= redFlagThreshold;
     if (pos && neg) {
       g.bucket = "observation";
       g.reason = "mixed";
     } else if (pos) {
-      g.bucket = "skill-candidate";
+      g.bucket = "candidate";
     } else if (neg) {
       g.bucket = "red-flag";
     } else {
@@ -2724,7 +2752,7 @@ function distillFeedback(ctx, opts = {}) {
       g.reason = "insufficient";
     }
   }
-  const order = { "skill-candidate": 0, "red-flag": 1, observation: 2 };
+  const order = { candidate: 0, "red-flag": 1, observation: 2 };
   return result.sort((a, b) => order[a.bucket] - order[b.bucket] || `${a.endpoint}/${a.kind}`.localeCompare(`${b.endpoint}/${b.kind}`));
 }
 function approvalStats(ctx) {
@@ -2779,18 +2807,18 @@ function approvalStats(ctx) {
 }
 function runRetrospective(ctx, opts = {}) {
   const groups = distillFeedback(ctx, opts);
-  const candidates = groups.filter((g) => g.bucket === "skill-candidate").length;
+  const candidates = groups.filter((g) => g.bucket === "candidate").length;
   const redFlags = groups.filter((g) => g.bucket === "red-flag").length;
   const cli = ctx.config.cli;
   const guidance = [];
   if (candidates > 0) {
     guidance.push(
-      `\u53D1\u73B0 ${candidates} \u4E2A skill \u5019\u9009\u7EC4:\u4F9D\u636E\u5176 evidence(\u8DEF\u5F84+comment)\u8D77\u8349 .claude/skills/<\u540D\u79F0>/SKILL.md,\u7136\u540E \`${cli} register-meta\` \u6CE8\u518C + \`${cli} submit --actor=<\u89D2\u8272> -- .claude/skills/<\u540D\u79F0>/SKILL.md\` \u9001\u4EBA\u5BA1;approved \u540E\u624D\u4F5C\u4E3A\u7ECF\u9A8C\u751F\u6548`
+      `\u53D1\u73B0 ${candidates} \u4E2A\u7ECF\u9A8C\u5019\u9009\u7EC4:\u5BF9\u6BCF\u7EC4\u4F9D\u636E evidence(\u8DEF\u5F84+comment)\u5224\u65AD\u8BE5\u6C89\u6DC0\u4E3A\u54EA\u4E00\u79CD,\u518D\u6309\u5BF9\u5E94\u8DEF\u5F84\u4EA7\u51FA\u2014\u2014\u8DE8\u4F1A\u8BDD\u53EF\u590D\u7528\u7684\u505A\u6CD5/\u6D41\u7A0B \u2192 skill(.claude/skills/<\u540D\u79F0>/SKILL.md;\`${cli} register-meta\` \u6CE8\u518C + \`${cli} submit --actor=<\u89D2\u8272> -- <\u8DEF\u5F84>\` \u9001\u4EBA\u5BA1,approved \u624D\u751F\u6548);\u5FC5\u987B\u59CB\u7EC8\u6210\u7ACB\u3001\u53EF grep \u7684\u786C\u7EA6\u675F \u2192 \u89C4\u5219(workbench.config.json \u7684 protocolLints \u5361\u70B9,\u6216\u5199\u5165 TECH.md / \u57FA\u7EBF\u7EA6\u5B9A);\u53EA\u5BF9\u7279\u5B9A\u89D2\u8272/\u9879\u76EE\u6709\u7528\u3001\u4E0D\u503C\u5F97\u5355\u72EC\u6210\u7BC7\u7684\u6559\u8BAD\u6216\u504F\u597D \u2192 \u8BB0\u5FC6(.claude/agent-memory/<\u89D2\u8272>/,\u66F4\u65B0 MEMORY.md \u7D22\u5F15)`
     );
   }
   if (redFlags > 0) {
     guidance.push(
-      `\u53D1\u73B0 ${redFlags} \u4E2A Red Flags \u7EC4:\u8D1F\u4F8B comment \u5373\u7D20\u6750,\u5199\u5165\u5BF9\u5E94 skill \u8349\u7A3F\u7684 Red Flags \u7AE0\u8282;\u80FD\u673A\u5668\u67E5\u7684\u7EA6\u5B9A\u5E94\u964D\u7EA7\u4E3A protocolLints \u5361\u70B9(workbench.config.json)`
+      `\u53D1\u73B0 ${redFlags} \u4E2A Red Flags \u7EC4:\u8D1F\u4F8B comment \u5373\u300C\u522B\u518D\u72AF\u300D\u7684\u7D20\u6750,\u540C\u6837\u4E09\u9009\u4E00\u2014\u2014\u80FD\u673A\u5668\u67E5 \u2192 protocolLints \u5361\u70B9;\u8DE8\u4F1A\u8BDD\u901A\u7528\u5751 \u2192 \u5199\u8FDB\u5BF9\u5E94 skill \u7684 Red Flags \u7AE0\u8282;\u89D2\u8272\u4E13\u5C5E\u5751 \u2192 \u8BB0\u5FC6`
     );
   }
   if (guidance.length === 0) {
@@ -2821,11 +2849,11 @@ function exportEventLog(ctx) {
     files: [`${ctx.config.dataDir}/events.jsonl`, `${ctx.config.dataDir}/feedback.jsonl`]
   };
 }
-var SKILL_CANDIDATE_THRESHOLD, RED_FLAG_THRESHOLD, AUTO_ACTOR_WEIGHT;
+var CANDIDATE_THRESHOLD, RED_FLAG_THRESHOLD, AUTO_ACTOR_WEIGHT;
 var init_retro_command = __esm({
   "core/commands/retro.command.ts"() {
     "use strict";
-    SKILL_CANDIDATE_THRESHOLD = 3;
+    CANDIDATE_THRESHOLD = 3;
     RED_FLAG_THRESHOLD = 2;
     AUTO_ACTOR_WEIGHT = 0.5;
   }
@@ -3377,6 +3405,7 @@ var init_init_command = __esm({
 var core_exports = {};
 __export(core_exports, {
   AUTO_ACTOR_WEIGHT: () => AUTO_ACTOR_WEIGHT,
+  CANDIDATE_THRESHOLD: () => CANDIDATE_THRESHOLD,
   CONFIG_FILENAME: () => CONFIG_FILENAME,
   CURRENT_SCHEMA_VERSION: () => CURRENT_SCHEMA_VERSION,
   DEFAULT_KIND_REGISTRY: () => DEFAULT_KIND_REGISTRY,
@@ -3388,7 +3417,6 @@ __export(core_exports, {
   PM_KINDS: () => PM_KINDS,
   RED_FLAG_THRESHOLD: () => RED_FLAG_THRESHOLD,
   SCHEMA_V1: () => SCHEMA_V1,
-  SKILL_CANDIDATE_THRESHOLD: () => SKILL_CANDIDATE_THRESHOLD,
   WORKBENCH_BIN: () => WORKBENCH_BIN,
   WORKBENCH_DIR: () => WORKBENCH_DIR,
   addTaskInput: () => addTaskInput,
@@ -3615,7 +3643,7 @@ function buildMcpServer(ctx) {
   server.registerTool(
     "wb_retro",
     {
-      description: "retrospective \u8BC1\u636E\u5305:\u53CD\u9988\u534A\u8870\u671F\u52A0\u6743\u63D0\u70BC(skill \u5019\u9009/Red Flags/\u89C2\u5BDF)+ \u5BA1\u6279\u541E\u5410\u62A5\u8868\u3002\u4F9D\u636E evidence \u8D77\u8349 .claude/skills/<\u540D\u79F0>/SKILL.md \u540E register-meta + submit \u9001\u4EBA\u5BA1,approved \u624D\u751F\u6548",
+      description: "retrospective \u8BC1\u636E\u5305:\u53CD\u9988\u534A\u8870\u671F\u52A0\u6743\u63D0\u70BC(\u7ECF\u9A8C\u5019\u9009/Red Flags/\u89C2\u5BDF)+ \u5BA1\u6279\u541E\u5410\u62A5\u8868\u3002\u4F9D\u636E evidence \u5224\u65AD\u6BCF\u7EC4\u8BE5\u6C89\u6DC0\u4E3A\u54EA\u4E00\u79CD\u518D\u4EA7\u51FA:\u53EF\u590D\u7528\u505A\u6CD5\u2192skill(.claude/skills/<\u540D\u79F0>/SKILL.md,register-meta+submit \u9001\u4EBA\u5BA1,approved \u624D\u751F\u6548);\u80FD\u673A\u5668\u67E5\u7684\u786C\u7EA6\u675F\u2192\u89C4\u5219(workbench.config.json \u7684 protocolLints);\u89D2\u8272\u4E13\u5C5E\u6559\u8BAD\u2192\u8BB0\u5FC6(.claude/agent-memory/<\u89D2\u8272>/)\u3002\u89C1 report.guidance",
       inputSchema: { module: z.string().optional() }
     },
     async (args) => json(runRetrospective(ctx, args))
@@ -3799,7 +3827,7 @@ data: {"cursor":${cursor2}}
   app.get("/api/skill-candidates", async () => {
     const report = runRetrospective(ctx);
     return {
-      groups: report.groups.filter((g) => g.bucket === "skill-candidate" || g.bucket === "red-flag"),
+      groups: report.groups.filter((g) => g.bucket === "candidate" || g.bucket === "red-flag"),
       candidates: report.candidates,
       redFlags: report.redFlags,
       halfLifeDays: report.halfLifeDays,
@@ -4411,7 +4439,7 @@ async function runCommand(ctx, command, a) {
     }
     case "scan": {
       const s = scanArtifacts(ctx, a.actor || "system");
-      console.log(chalk.green(`\u2705 \u626B\u63CF\u5B8C\u6210:\u65B0\u767B\u8BB0 ${s.registered},\u5185\u5BB9\u5237\u65B0 ${s.refreshed},\u65B0\u589E\u8FB9 ${s.edges},\u6392\u9664\u5143\u4EA7\u7269 ${s.skipped.length}`));
+      console.log(chalk.green(`\u2705 \u626B\u63CF\u5B8C\u6210:\u65B0\u767B\u8BB0 ${s.registered},\u5185\u5BB9\u5237\u65B0 ${s.refreshed},\u91CD\u6302\u5750\u6807 ${s.remapped},\u65B0\u589E\u8FB9 ${s.edges},\u6392\u9664\u5143\u4EA7\u7269 ${s.skipped.length}`));
       break;
     }
     case "move": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dawipong/opcflow",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "description": "Spec-anchored, drift-enforced execution layer for AI coding agents — artifact DAG, trust gates, multi-role pipeline; generates agents/MCP/hooks for Claude Code, Codex, OpenCode & Cursor.",
   "license": "MIT",
   "author": "Dawi",

package/web/dist/assets/{abnfDiagram-VRR7QNED-DctuL32B.js → abnfDiagram-VRR7QNED-B5Ymr02_.js}

@@ -1 +1 @@
-import{m as e}from"./src-W4W9ly-y.js";import{n as t}from"./chunk-Y2CYZVJY-DsF7k-Jl.js";import"./chunk-WYO6CB5R-BaRUDpDl.js";import"./index-DVXPy-Pj.js";import{m as n,t as r}from"./mermaid-parser.core-CAAH_OSq.js";import{n as i,r as a,t as o}from"./chunk-MOJQB5TN-Y0jSjq4a.js";import{t as s}from"./chunk-JWPE2WC7-DVXcaiue.js";var c=n().RailroadAbnf.parser.LangiumParser,l=t(e=>{let t=e.alternatives.map(u);return t.length===1?t[0]:{type:`choice`,alternatives:t}},`transformAlternation`),u=t(e=>{let t=e.elements.map(f);return t.length===1?t[0]:{type:`sequence`,elements:t}},`transformConcatenation`),d=t(e=>{if(e.includes(`*`)){let[t,n]=e.split(`*`);return{min:t?parseInt(t,10):0,max:n?parseInt(n,10):1/0}}let t=parseInt(e,10);return{min:t,max:t}},`parseRepeat`),f=t(e=>{let t=p(e.primary);if(!e.repeat)return t;let{min:n,max:r}=d(e.repeat);return n===0&&r===1?{type:`optional`,element:t}:{type:`repetition`,element:t,min:n,max:r}},`transformElement`),p=t(e=>{switch(e.$type){case`AbnfStringLiteral`:return{type:`terminal`,value:e.value};case`AbnfNumVal`:return{type:`terminal`,value:e.value};case`AbnfRuleName`:return{type:`nonterminal`,name:e.name};case`AbnfGroup`:return l(e.element);case`AbnfOptionalGroup`:return{type:`optional`,element:l(e.element)};default:throw Error(`Unsupported ABNF primary node: ${e.$type}`)}},`transformPrimary`),m=t(e=>({name:e.name,definition:l(e.definition)}),`transformRule`),h=t(e=>{s(e,o),e.title&&o.setTitle(e.title),e.rules.map(e=>o.addRule(m(e)))},`populateDb`),g={parser:{parse:t(t=>{o.clear(),e.debug(`[ABNF Parser] Starting Langium parse`);let n=c.parse(t);if(n.lexerErrors.length>0||n.parserErrors.length>0)throw new r(n);let i=n.value;e.debug(`[ABNF Parser] Parsed rules:`,i.rules.length),h(i),e.debug(`[ABNF Parser] Parse complete`)},`parse`),parser:{yy:o}},db:o,renderer:a,styles:i};export{g as diagram};
+import{m as e}from"./src-W4W9ly-y.js";import{n as t}from"./chunk-Y2CYZVJY-DsF7k-Jl.js";import"./chunk-WYO6CB5R-BaRUDpDl.js";import"./index-D02utICb.js";import{m as n,t as r}from"./mermaid-parser.core-CAAH_OSq.js";import{n as i,r as a,t as o}from"./chunk-MOJQB5TN-BJQ2jnj0.js";import{t as s}from"./chunk-JWPE2WC7-DVXcaiue.js";var c=n().RailroadAbnf.parser.LangiumParser,l=t(e=>{let t=e.alternatives.map(u);return t.length===1?t[0]:{type:`choice`,alternatives:t}},`transformAlternation`),u=t(e=>{let t=e.elements.map(f);return t.length===1?t[0]:{type:`sequence`,elements:t}},`transformConcatenation`),d=t(e=>{if(e.includes(`*`)){let[t,n]=e.split(`*`);return{min:t?parseInt(t,10):0,max:n?parseInt(n,10):1/0}}let t=parseInt(e,10);return{min:t,max:t}},`parseRepeat`),f=t(e=>{let t=p(e.primary);if(!e.repeat)return t;let{min:n,max:r}=d(e.repeat);return n===0&&r===1?{type:`optional`,element:t}:{type:`repetition`,element:t,min:n,max:r}},`transformElement`),p=t(e=>{switch(e.$type){case`AbnfStringLiteral`:return{type:`terminal`,value:e.value};case`AbnfNumVal`:return{type:`terminal`,value:e.value};case`AbnfRuleName`:return{type:`nonterminal`,name:e.name};case`AbnfGroup`:return l(e.element);case`AbnfOptionalGroup`:return{type:`optional`,element:l(e.element)};default:throw Error(`Unsupported ABNF primary node: ${e.$type}`)}},`transformPrimary`),m=t(e=>({name:e.name,definition:l(e.definition)}),`transformRule`),h=t(e=>{s(e,o),e.title&&o.setTitle(e.title),e.rules.map(e=>o.addRule(m(e)))},`populateDb`),g={parser:{parse:t(t=>{o.clear(),e.debug(`[ABNF Parser] Starting Langium parse`);let n=c.parse(t);if(n.lexerErrors.length>0||n.parserErrors.length>0)throw new r(n);let i=n.value;e.debug(`[ABNF Parser] Parsed rules:`,i.rules.length),h(i),e.debug(`[ABNF Parser] Parse complete`)},`parse`),parser:{yy:o}},db:o,renderer:a,styles:i};export{g as diagram};