flower-trellis 0.4.12-beta.2 → 0.5.0-beta.0

package/enhancements/0.6/.agents/skills/trellis-push/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: trellis-push
-description: "按确认的精确文件范围提交普通变更或完成已就绪的 merge commit；多仓计划可包含已展示的本地生成命令，并在普通推送后同步当前任务进度。"
+description: "按确认的精确文件范围提交普通变更或完成已就绪的 merge commit；多仓计划可包含已展示的本地生成命令，并在普通推送后同步当前任务记录与进度。"
 ---
 
 # Trellis Push
@@ -11,6 +11,7 @@ description: "按确认的精确文件范围提交普通变更或完成已就绪
 
 - 普通模式默认 `commit + push`。
 - 普通多仓计划可以包含本地确定性生成命令；生成后没有新增计划外文件时沿用同一次确认。
+- 普通模式把当前任务产物与更新后的 `task.json` 纳入同一次确认下的独立任务记录提交。
 - 用户明确要求“只提交不推送”时使用 `commit-only`。
 - auto-loop 可调用内部 `commit-only`，但必须传入已经校验过的 exact files 与 commit message；本 skill 只执行该提交。
 - 不发起、终止或解决分支合并；只允许普通模式完成已经开始、冲突已清零且索引完全可归属的 merge commit。
@@ -46,7 +47,13 @@ python3 ./.trellis/scripts/task.py current --source || true
 python3 ./.trellis/scripts/task_progress.py status --json || true
 ```
 
-无活动任务时仍可提交相关代码，但不生成任务进度。存在活动任务时，结合 `brief.md`、`implement.md`、当前 diff 与本轮执行范围生成一行语义进度；不得从旧进度推断 Git 动作。
+存在活动任务时，必须额外获取文件级任务状态：
+
+```bash
+git status --short --untracked-files=all -- <task-dir>
+```
+
+不得把默认 `git status --short` 可能返回的 `?? <task-dir>/` 折叠目录当成 exact file、展示条目或 pathspec。无活动任务时仍可提交相关代码，但不生成任务进度。存在活动任务时，结合 `brief.md`、`implement.md`、当前 diff 与本轮执行范围生成一行语义进度；同时识别当前任务目录中已存在且可归属的 dirty/untracked 产物，供 Step 5 生成任务记录 exact files。不得从旧进度推断 Git 动作。
 
 ## Step 2：预检与文件归属
 
@@ -75,11 +82,13 @@ git log @{u}..HEAD --oneline 2>/dev/null || true
 - 无法确定 planned file 是否属于当前请求或活动任务。
 - 内部 `commit-only` 发现 staged 区非空。
 
-文件分为两组：
+业务 Git 文件分为两组；普通模式的当前任务记录 exact files 按下方独立提交规则处理：
 
 - `planned`：本轮明确归属且准备提交的 exact files。
 - `retained`：当前存在、但本次明确不提交并保持原状的 dirty paths，包含计划外 untracked、unstaged、staged 文件。clean files 不进入该集合。
 
+普通模式存在活动任务时，当前任务目录中已存在且可归属的 dirty/untracked 产物不进入业务 `planned`，也不进入 `retained`；它们与预计由 helper 更新的 `<task-dir>/task.json` 组成 Step 5 的任务记录 exact files。其他任务目录和无法归属当前任务的文件仍属于 `retained` 或风险，不得顺带提交。
+
 普通 `PUSH` 需要在仓库间运行本地生成命令时，首次计划同时展示命令、工作目录和后续仓预计 exact files。仅在后续仓没有 retained dirty 时使用；命令必须本地、可重复且无外部副作用。
 
 `retained` 只是内部集合名。用户可见输出统一写“保留未提交的变更（dirty）”，并逐项标注 `[untracked]`、`[unstaged]`、`[staged]`。unknown ahead、branch/upstream 异常、归属不确定等真正需要处理的事项单独进入“风险”区；普通 retained dirty 不默认视为阻塞。
@@ -121,6 +130,9 @@ Push：<执行 / 跳过（commit-only）>
 ### 风险（仅数量大于 0 时显示）
 - <unknown ahead / branch-upstream / attribution risk>
 
+[任务记录（仅普通模式且存在活动任务时显示）：`chore(task): update <task-name> progress` · <N> 个文件]
+[仓库：<repository-name> · 分支：`<branch>` -> `<upstream>`]
+[计划提交：<当前任务 exact files 或分组摘要>]
 任务进度：completed=<...> | partial=<...> | next=<...>
 执行：<commit -> push -> progress commit -> progress push>
 
@@ -131,6 +143,7 @@ Push：<执行 / 跳过（commit-only）>
 
 - 单仓 `planned` 不超过 8 个文件时完整列出。
 - 超过 8 个时按目录归组，最多 12 行；用户要求展开时展示同一 exact set。
+- 顶部仓库/commit/file 总数包含独立任务记录提交所在 Git root、该提交及其 exact files；任务记录文件使用相同的 8 文件展示阈值和展开规则。
 - 保留未提交的变更始终逐项标注 Git 状态；真正风险在独立“风险”区逐项展示。
 - 无活动任务或 `commit-only` 时省略进度动作。
 - 不重复展示检查结果、规范复核、归档或其他阶段信息。
@@ -214,7 +227,7 @@ git push origin <current-branch>
 
 写入前确认：
 
-- 当前任务 `task.json` 在业务提交结束后没有计划外 dirty 内容。
+- 当前任务 exact files 与首次确认的路径集合一致，没有新增当前任务路径或无法归属的 dirty 内容。
 - 父仓分支、upstream 和冲突状态安全。
 - 推送不会携带无法归属的历史 ahead commits。
 
@@ -227,15 +240,15 @@ python3 ./.trellis/scripts/task_progress.py write \
   --json
 ```
 
-然后只提交并推送当前任务 `task.json`：
+然后只提交并推送首次确认的当前任务 exact files；该集合包含 helper 更新后的 `task.json`，以及首次计划时已存在且可归属的当前任务 dirty/untracked 产物：
 
 ```bash
-git add -- <task-dir>/task.json
-git commit --only -m "chore(task): update <task-name> progress" -- <task-dir>/task.json
+git add -- <current-task-exact-files>
+git commit --only -m "chore(task): update <task-name> progress" -- <current-task-exact-files>
 git push origin <current-branch>
 ```
 
-该动作属于用户已确认的普通 push 计划，不增加第二次确认。提交后必须验证 commit 只包含该 `task.json`。如果写入、提交或推送失败，不回滚已成功的业务 Git 动作，并单独报告进度同步失败。
+该动作属于用户已确认的普通 push 计划，不增加第二次确认。提交后必须验证 commit 只包含首次确认的当前任务 exact files；其他任务和无关 dirty/staged 文件保持原状。如果写入、提交或推送失败，不回滚已成功的业务 Git 动作，并单独报告进度同步失败。
 
 ## Step 6：结果
 
@@ -257,6 +270,7 @@ git push origin <current-branch>
 ### 任务进度
 
 状态：<✓ 已同步 · `<progress-hash>` / · 已跳过 / ❌ 同步失败>
+记录：<N> 个当前任务文件
 进度：completed=<...> | partial=<...> | next=<...>
 [失败时追加：原因和恢复动作]
 
@@ -271,6 +285,7 @@ git push origin <current-branch>
 ## 禁止事项
 
 - 扩大到计划外文件或要求清理无关工作区。
+- 把普通 push 中可归属当前活动任务的规划产物列为 retained，并以“finish-work 归档时再入库”为由延后首次记录。
 - 执行首次计划未展示的生成命令，或生成计划外文件后仍沿用旧确认。
 - 用任务进度决定是否推送代码。
 - 在本 skill 内发起、终止、解决冲突或改变分支合并目标；只允许完成已就绪的 merge commit。

package/enhancements/0.6/.claude/skills/trellis-push/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: trellis-push
-description: "按确认的精确文件范围提交普通变更或完成已就绪的 merge commit；多仓计划可包含已展示的本地生成命令，并在普通推送后同步当前任务进度。"
+description: "按确认的精确文件范围提交普通变更或完成已就绪的 merge commit；多仓计划可包含已展示的本地生成命令，并在普通推送后同步当前任务记录与进度。"
 ---
 
 # Trellis Push
@@ -11,6 +11,7 @@ description: "按确认的精确文件范围提交普通变更或完成已就绪
 
 - 普通模式默认 `commit + push`。
 - 普通多仓计划可以包含本地确定性生成命令；生成后没有新增计划外文件时沿用同一次确认。
+- 普通模式把当前任务产物与更新后的 `task.json` 纳入同一次确认下的独立任务记录提交。
 - 用户明确要求“只提交不推送”时使用 `commit-only`。
 - auto-loop 可调用内部 `commit-only`，但必须传入已经校验过的 exact files 与 commit message；本 skill 只执行该提交。
 - 不发起、终止或解决分支合并；只允许普通模式完成已经开始、冲突已清零且索引完全可归属的 merge commit。
@@ -46,7 +47,13 @@ python3 ./.trellis/scripts/task.py current --source || true
 python3 ./.trellis/scripts/task_progress.py status --json || true
 ```
 
-无活动任务时仍可提交相关代码，但不生成任务进度。存在活动任务时，结合 `brief.md`、`implement.md`、当前 diff 与本轮执行范围生成一行语义进度；不得从旧进度推断 Git 动作。
+存在活动任务时，必须额外获取文件级任务状态：
+
+```bash
+git status --short --untracked-files=all -- <task-dir>
+```
+
+不得把默认 `git status --short` 可能返回的 `?? <task-dir>/` 折叠目录当成 exact file、展示条目或 pathspec。无活动任务时仍可提交相关代码，但不生成任务进度。存在活动任务时，结合 `brief.md`、`implement.md`、当前 diff 与本轮执行范围生成一行语义进度；同时识别当前任务目录中已存在且可归属的 dirty/untracked 产物，供 Step 5 生成任务记录 exact files。不得从旧进度推断 Git 动作。
 
 ## Step 2：预检与文件归属
 
@@ -75,11 +82,13 @@ git log @{u}..HEAD --oneline 2>/dev/null || true
 - 无法确定 planned file 是否属于当前请求或活动任务。
 - 内部 `commit-only` 发现 staged 区非空。
 
-文件分为两组：
+业务 Git 文件分为两组；普通模式的当前任务记录 exact files 按下方独立提交规则处理：
 
 - `planned`：本轮明确归属且准备提交的 exact files。
 - `retained`：当前存在、但本次明确不提交并保持原状的 dirty paths，包含计划外 untracked、unstaged、staged 文件。clean files 不进入该集合。
 
+普通模式存在活动任务时，当前任务目录中已存在且可归属的 dirty/untracked 产物不进入业务 `planned`，也不进入 `retained`；它们与预计由 helper 更新的 `<task-dir>/task.json` 组成 Step 5 的任务记录 exact files。其他任务目录和无法归属当前任务的文件仍属于 `retained` 或风险，不得顺带提交。
+
 普通 `PUSH` 需要在仓库间运行本地生成命令时，首次计划同时展示命令、工作目录和后续仓预计 exact files。仅在后续仓没有 retained dirty 时使用；命令必须本地、可重复且无外部副作用。
 
 `retained` 只是内部集合名。用户可见输出统一写“保留未提交的变更（dirty）”，并逐项标注 `[untracked]`、`[unstaged]`、`[staged]`。unknown ahead、branch/upstream 异常、归属不确定等真正需要处理的事项单独进入“风险”区；普通 retained dirty 不默认视为阻塞。
@@ -121,6 +130,9 @@ Push：<执行 / 跳过（commit-only）>
 ### 风险（仅数量大于 0 时显示）
 - <unknown ahead / branch-upstream / attribution risk>
 
+[任务记录（仅普通模式且存在活动任务时显示）：`chore(task): update <task-name> progress` · <N> 个文件]
+[仓库：<repository-name> · 分支：`<branch>` -> `<upstream>`]
+[计划提交：<当前任务 exact files 或分组摘要>]
 任务进度：completed=<...> | partial=<...> | next=<...>
 执行：<commit -> push -> progress commit -> progress push>
 
@@ -131,6 +143,7 @@ Push：<执行 / 跳过（commit-only）>
 
 - 单仓 `planned` 不超过 8 个文件时完整列出。
 - 超过 8 个时按目录归组，最多 12 行；用户要求展开时展示同一 exact set。
+- 顶部仓库/commit/file 总数包含独立任务记录提交所在 Git root、该提交及其 exact files；任务记录文件使用相同的 8 文件展示阈值和展开规则。
 - 保留未提交的变更始终逐项标注 Git 状态；真正风险在独立“风险”区逐项展示。
 - 无活动任务或 `commit-only` 时省略进度动作。
 - 不重复展示检查结果、规范复核、归档或其他阶段信息。
@@ -214,7 +227,7 @@ git push origin <current-branch>
 
 写入前确认：
 
-- 当前任务 `task.json` 在业务提交结束后没有计划外 dirty 内容。
+- 当前任务 exact files 与首次确认的路径集合一致，没有新增当前任务路径或无法归属的 dirty 内容。
 - 父仓分支、upstream 和冲突状态安全。
 - 推送不会携带无法归属的历史 ahead commits。
 
@@ -227,15 +240,15 @@ python3 ./.trellis/scripts/task_progress.py write \
   --json
 ```
 
-然后只提交并推送当前任务 `task.json`：
+然后只提交并推送首次确认的当前任务 exact files；该集合包含 helper 更新后的 `task.json`，以及首次计划时已存在且可归属的当前任务 dirty/untracked 产物：
 
 ```bash
-git add -- <task-dir>/task.json
-git commit --only -m "chore(task): update <task-name> progress" -- <task-dir>/task.json
+git add -- <current-task-exact-files>
+git commit --only -m "chore(task): update <task-name> progress" -- <current-task-exact-files>
 git push origin <current-branch>
 ```
 
-该动作属于用户已确认的普通 push 计划，不增加第二次确认。提交后必须验证 commit 只包含该 `task.json`。如果写入、提交或推送失败，不回滚已成功的业务 Git 动作，并单独报告进度同步失败。
+该动作属于用户已确认的普通 push 计划，不增加第二次确认。提交后必须验证 commit 只包含首次确认的当前任务 exact files；其他任务和无关 dirty/staged 文件保持原状。如果写入、提交或推送失败，不回滚已成功的业务 Git 动作，并单独报告进度同步失败。
 
 ## Step 6：结果
 
@@ -257,6 +270,7 @@ git push origin <current-branch>
 ### 任务进度
 
 状态：<✓ 已同步 · `<progress-hash>` / · 已跳过 / ❌ 同步失败>
+记录：<N> 个当前任务文件
 进度：completed=<...> | partial=<...> | next=<...>
 [失败时追加：原因和恢复动作]
 
@@ -271,6 +285,7 @@ git push origin <current-branch>
 ## 禁止事项
 
 - 扩大到计划外文件或要求清理无关工作区。
+- 把普通 push 中可归属当前活动任务的规划产物列为 retained，并以“finish-work 归档时再入库”为由延后首次记录。
 - 执行首次计划未展示的生成命令，或生成计划外文件后仍沿用旧确认。
 - 用任务进度决定是否推送代码。
 - 在本 skill 内发起、终止、解决冲突或改变分支合并目标；只允许完成已就绪的 merge commit。

package/enhancements/0.6/overrides/transforms/content/brainstorm-auto-task-create.md

@@ -0,0 +1,15 @@
+If no task exists yet, choose the creator from the authorization source.
+
+For inferred high-confidence complex implementation intent, create an auto-routed planning task:
+
+```bash
+python3 ./.trellis/scripts/task_intent.py create --title "<short task title>" --slug <slug>
+```
+
+For explicit user-requested task planning or a manually maintained task, use:
+
+```bash
+python3 ./.trellis/scripts/task.py create "<short task title>" --slug <slug>
+```
+
+Use a concise title from the user's request and a slug without a date prefix. Both paths create the default `prd.md`; update it with the current understanding before asking follow-up questions.

package/enhancements/0.6/overrides/transforms/content/brainstorm-planning-authorization.md

@@ -0,0 +1 @@
+Use this skill after the user explicitly requests task planning, or after clear complex implementation intent authorizes creating a planning workspace. This authorization covers planning only; it never authorizes `task.py start` or implementation.

package/enhancements/0.6/overrides/transforms/content/claude-session-start-no-task.py

@@ -0,0 +1,6 @@
+        return (
+            "Status: NO ACTIVE TASK\n"
+            "Next-Action: Infer discuss, inspect, direct_edit, task_plan, or workflow_action. "
+            "Proceed with high-confidence reversible routing; ask only for material ambiguity "
+            "or independent safety gates."
+        )

package/enhancements/0.6/overrides/transforms/content/codex-session-start-no-task.py

@@ -0,0 +1,6 @@
+        return (
+            "Status: NO ACTIVE TASK\n"
+            "Next: Infer discuss, inspect, direct_edit, task_plan, or workflow_action. "
+            "Proceed with high-confidence reversible routing; ask only for material ambiguity "
+            "or independent safety gates."
+        )

package/enhancements/0.6/overrides/transforms/content/start-no-task-routing.md

@@ -0,0 +1 @@
+- **No active task** → infer `discuss`, `inspect`, `direct_edit`, `task_plan`, or `workflow_action`. Proceed with high-confidence reversible routing; inferred complex implementation uses `task_intent.py create`, explicit task planning uses `task.py create`, and material ambiguity or independent safety boundaries still require one focused question.

package/enhancements/0.6/overrides/transforms/content/workflow-create-task-command.md

@@ -0,0 +1,11 @@
+For inferred high-confidence complex implementation intent, preserve request scope and the dirty baseline:
+
+```bash
+python3 ./.trellis/scripts/task_intent.py create --title "<task title>" --slug <name>
+```
+
+For explicit user-requested task planning or a manually maintained task, use the ordinary creator:
+
+```bash
+python3 ./.trellis/scripts/task.py create "<task title>" --slug <name>
+```

package/enhancements/0.6/overrides/transforms/content/workflow-create-task-rule.md

@@ -0,0 +1 @@
+Create the task directory after explicit task intent or high-confidence complex implementation intent authorizes planning. Auto-routed creation should use `task_intent.py create` so request scope and the pre-planning dirty baseline are recorded. The task remains `planning`, writes `task.json`, creates a default `prd.md`, and targets the current session when identity is available:

package/enhancements/0.6/overrides/transforms/content/workflow-customization-intent-invariant.md

@@ -0,0 +1 @@
+- No active task must infer the current request intent first; high-confidence reversible routing proceeds directly, while material ambiguity and independent safety boundaries still require confirmation.

package/enhancements/0.6/overrides/transforms/content/workflow-no-task-body.md

@@ -0,0 +1,3 @@
+No active task. Infer the current request intent before acting.
+Handle `discuss` and `inspect` silently. For `workflow_action`, load the named Trellis capability directly. For non-destructive `direct_edit`, state once that task/progress will not be recorded and proceed.
+For high-confidence complex implementation, create an auto-routed planning task through `task_intent.py create`, show one non-blocking switch hint, and enter `trellis-brainstorm`. Ask only for material ambiguity or independent safety gates.

package/enhancements/0.6/overrides/transforms/content/workflow-phase-index-create-task.md

@@ -0,0 +1 @@
+- 1.0 Create task `[required · once]` (when `task_plan` is explicit or inferred from clear complex implementation intent)

package/enhancements/0.6/overrides/transforms/content/workflow-phase-one-goal.md

@@ -0,0 +1 @@
+Goal: infer the request intent, enter planning automatically when authorized by explicit or high-confidence complex implementation intent, and produce the artifacts required before implementation.

package/enhancements/0.6/overrides/transforms/content/workflow-phase-summary.md

@@ -0,0 +1 @@
+Phase 1: Plan    → infer intent, take the authorized reversible next step, and write planning artifacts when needed

package/enhancements/0.6/overrides/transforms/content/workflow-request-triage.md

@@ -0,0 +1,4 @@
+- Infer `discuss`, `inspect`, `direct_edit`, `task_plan`, or `workflow_action` from the current request, its scope, risk, side effects, active-task state, and the latest explicit switch. Do not classify from one keyword alone.
+- High-confidence reversible steps proceed without a mechanical task-creation question. Explicit complex implementation intent authorizes creating a planning task and entering `trellis-brainstorm`; it does not authorize `task.py start` or implementation.
+- Ask one question only when ambiguity changes material side effects, or when destructive, production, database, credential, external-system, or permission boundaries require confirmation.
+- The latest explicit `走任务` / `不要任务` / `先讨论` / `直接做` / `先别做` style instruction wins for the current request. New unrelated requests return to automatic inference.

package/enhancements/0.6/overrides/transforms/intent-routing.json

@@ -0,0 +1,248 @@
+{
+  "schemaVersion": 1,
+  "id": "intent-routing",
+  "aliases": [
+    "workflow-enhancement",
+    "task-intent",
+    "intent-routing"
+  ],
+  "operations": [
+    {
+      "id": "workflow-phase-summary",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-phase-summary.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-phase-summary.md"
+      }
+    },
+    {
+      "id": "workflow-request-triage",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-request-triage.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-request-triage.md"
+      }
+    },
+    {
+      "id": "workflow-no-task-body",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-no-task-body.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-no-task-body.md"
+      }
+    },
+    {
+      "id": "workflow-phase-index-create-task",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-phase-index-create-task.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-phase-index-create-task.md"
+      }
+    },
+    {
+      "id": "workflow-phase-one-goal",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-phase-one-goal.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-phase-one-goal.md"
+      }
+    },
+    {
+      "id": "workflow-create-task-rule",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-create-task-rule.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-create-task-rule.md"
+      }
+    },
+    {
+      "id": "workflow-create-task-command",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-create-task-command.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-create-task-command.md"
+      }
+    },
+    {
+      "id": "workflow-customization-intent-invariant",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "workflow", "path": ".trellis/workflow.md" }
+      ],
+      "selector": {
+        "source": "matches/workflow-customization-intent-invariant.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/workflow-customization-intent-invariant.md"
+      }
+    },
+    {
+      "id": "codex-session-start-no-task",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "hook", "path": ".codex/hooks/session-start.py", "markerStyle": "hash" }
+      ],
+      "selector": {
+        "source": "matches/codex-session-start-no-task.py",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/codex-session-start-no-task.py"
+      }
+    },
+    {
+      "id": "claude-session-start-no-task",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "hook", "path": ".claude/hooks/session-start.py", "markerStyle": "hash" }
+      ],
+      "selector": {
+        "source": "matches/claude-session-start-no-task.py",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/claude-session-start-no-task.py"
+      }
+    },
+    {
+      "id": "start-no-task-routing",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "skill", "path": ".agents/skills/trellis-start/SKILL.md" },
+        { "kind": "command", "path": ".agent/workflows/start.md" },
+        { "kind": "command", "path": ".claude/commands/trellis/start.md" },
+        { "kind": "command", "path": ".codebuddy/commands/trellis/start.md" },
+        { "kind": "command", "path": ".cursor/commands/trellis/start.md" },
+        { "kind": "command", "path": ".devin/workflows/trellis-start.md" },
+        { "kind": "command", "path": ".github/prompts/start.prompt.md" },
+        { "kind": "command", "path": ".kilocode/workflows/start.md" },
+        { "kind": "command", "path": ".opencode/commands/trellis/start.md" },
+        { "kind": "command", "path": ".pi/prompts/trellis-start.md" },
+        { "kind": "command", "path": ".qoder/commands/trellis/start.md" },
+        { "kind": "skill", "path": ".reasonix/skills/trellis-start/SKILL.md" },
+        { "kind": "command", "path": ".zcode/commands/trellis/start.md" }
+      ],
+      "selector": {
+        "source": "matches/start-no-task-routing.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/start-no-task-routing.md"
+      }
+    },
+    {
+      "id": "brainstorm-planning-authorization",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "skill", "path": ".agent/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".agents/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".claude/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".codebuddy/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".cursor/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".devin/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".factory/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".github/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".kilocode/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".kiro/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".opencode/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".pi/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".qoder/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".reasonix/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".trae/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".zcode/skills/trellis-brainstorm/SKILL.md" }
+      ],
+      "selector": {
+        "source": "matches/brainstorm-planning-authorization.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/brainstorm-planning-authorization.md"
+      }
+    },
+    {
+      "id": "brainstorm-auto-task-create",
+      "operation": "replace",
+      "required": true,
+      "targets": [
+        { "kind": "skill", "path": ".agent/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".agents/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".claude/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".codebuddy/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".cursor/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".devin/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".factory/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".github/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".kilocode/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".kiro/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".opencode/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".pi/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".qoder/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".reasonix/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".trae/skills/trellis-brainstorm/SKILL.md" },
+        { "kind": "skill", "path": ".zcode/skills/trellis-brainstorm/SKILL.md" }
+      ],
+      "selector": {
+        "source": "matches/brainstorm-auto-task-create.md",
+        "expectedMatches": 1
+      },
+      "content": {
+        "source": "content/brainstorm-auto-task-create.md"
+      }
+    }
+  ]
+}

package/enhancements/0.6/overrides/transforms/matches/brainstorm-auto-task-create.md

@@ -0,0 +1,9 @@
+If no task exists yet, create one:
+
+```bash
+TASK_DIR=$(python3 ./.trellis/scripts/task.py create "<short task title>" --slug <slug>)
+```
+
+Use a concise title from the user's request. Use a slug without a date prefix. `task.py create` adds the `MM-DD-` directory prefix automatically.
+
+`task.py create` creates the default `prd.md`. Update that file with the current understanding before asking follow-up questions.

package/enhancements/0.6/overrides/transforms/matches/brainstorm-planning-authorization.md

@@ -0,0 +1 @@
+Use this skill only after task-creation consent has been given and the user is ready to enter Trellis planning.

package/enhancements/0.6/overrides/transforms/matches/claude-session-start-no-task.py

@@ -0,0 +1,6 @@
+        return (
+            "Status: NO ACTIVE TASK\n"
+            "Next-Action: Classify the current turn before creating any Trellis task. "
+            "Simple conversation / small task asks only whether this turn should create a Trellis task. "
+            "Complex task asks whether task creation and planning are allowed."
+        )

package/enhancements/0.6/overrides/transforms/matches/codex-session-start-no-task.py

@@ -0,0 +1,5 @@
+        return (
+            "Status: NO ACTIVE TASK\n"
+            "Next: Classify the current turn and ask for task-creation consent "
+            "before creating any Trellis task."
+        )

package/enhancements/0.6/overrides/transforms/matches/start-no-task-routing.md

@@ -0,0 +1 @@
+- **No active task** → classify first. For simple conversation / small task, ask only whether this turn should create a Trellis task. For complex work, ask whether you may create a Trellis task and enter planning. If the user says no, skip Trellis for this session.

package/enhancements/0.6/overrides/transforms/matches/workflow-create-task-command.md

@@ -0,0 +1,3 @@
+```bash
+python3 ./.trellis/scripts/task.py create "<task title>" --slug <name>
+```

package/enhancements/0.6/overrides/transforms/matches/workflow-create-task-rule.md

@@ -0,0 +1 @@
+Create the task directory only after task-creation consent. The command sets status to `planning`, writes `task.json`, creates a default `prd.md`, and auto-targets the new task when session identity is available:

package/enhancements/0.6/overrides/transforms/matches/workflow-customization-intent-invariant.md

@@ -0,0 +1 @@
+- No active task must triage first and ask for task-creation consent before creating a Trellis task.

package/enhancements/0.6/overrides/transforms/matches/workflow-no-task-body.md

@@ -0,0 +1,3 @@
+No active task. First classify the current turn and ask for task-creation consent before creating any Trellis task.
+Simple conversation / small task: ask only whether this turn should create a Trellis task. If the user says no, skip Trellis for this session.
+Complex task: ask the user if you can create a Trellis task and enter the planning phase. If the user says no, explain, clarify scope, or suggest a smaller split.

package/enhancements/0.6/overrides/transforms/matches/workflow-phase-index-create-task.md

@@ -0,0 +1 @@
+- 1.0 Create task `[required · once]` (only after task-creation consent)

package/enhancements/0.6/overrides/transforms/matches/workflow-phase-one-goal.md

@@ -0,0 +1 @@
+Goal: classify the request, get task-creation consent when a task is needed, and produce the planning artifacts required before implementation.

package/enhancements/0.6/overrides/transforms/matches/workflow-phase-summary.md

@@ -0,0 +1 @@
+Phase 1: Plan    → classify, get task-creation consent, then write planning artifacts

package/enhancements/0.6/overrides/transforms/matches/workflow-request-triage.md

@@ -0,0 +1,3 @@
+- Simple conversation or small task: ask only whether this turn should create a Trellis task. If the user says no, skip Trellis for this session.
+- Complex task: ask whether you may create a Trellis task and enter planning. If the user says no, do not do broad inline implementation; explain, clarify scope, or suggest a smaller split.
+- User approval to create a task is not approval to start implementation. Planning still happens first.

package/enhancements/0.6/overrides/workflow-states/planning-inline.md

@@ -3,6 +3,7 @@ HIGHEST PRIORITY SKILL-GARDEN STATE GUARD (planning-inline):
 Planning is not implementation permission.
 `trellis-brainstorm` is the default next action while requirements are still unclear.
 A created task or existing `prd.md` is not enough to start implementation.
+If the latest current-request switch says no task/direct edit, call `task_intent.py discard --task <current-task>` only for a task auto-created by intent routing; leave that task only on `status=discarded`. Keep manual or historical tasks unchanged and route the current request as untracked under the Active Task Scope Guard.
 Complete prd.md + required context first.
 If the active workflow later routes to sub-agent execution, required context includes real curated entries in both `implement.jsonl` and `check.jsonl`; the seed `_example` row alone is not ready.
 Before `task.py start`, use `trellis-task-brief` to refresh `brief.md` from the latest task artifacts and display it in chat for review.

package/enhancements/0.6/overrides/workflow-states/planning.md

@@ -3,6 +3,7 @@ HIGHEST PRIORITY SKILL-GARDEN STATE GUARD (planning):
 Planning is not implementation permission.
 `trellis-brainstorm` is the default next action while requirements are still unclear.
 A created task or existing `prd.md` is not enough to start implementation.
+If the latest current-request switch says no task/direct edit, call `task_intent.py discard --task <current-task>` only for a task auto-created by intent routing; leave that task only on `status=discarded`. Keep manual or historical tasks unchanged and route the current request as untracked under the Active Task Scope Guard.
 Complete prd.md + required context first.
 For sub-agent-dispatch platforms, required context includes real curated entries in both `implement.jsonl` and `check.jsonl`; the seed `_example` row alone is not ready.
 Before `task.py start`, use `trellis-task-brief` to refresh `brief.md` from the latest task artifacts and display it in chat for review.