@fitlab-ai/agent-infra 0.6.2-alpha.1 → 0.6.3

package/templates/.agents/skills/complete-task/config/verify.zh-CN.json

@@ -0,0 +1,48 @@
+{
+  "skill": "complete-task",
+  "checks": {
+    "task-meta": {
+      "required_fields": [
+        "id",
+        "type",
+        "workflow",
+        "status",
+        "created_at",
+        "updated_at",
+        "agent_infra_version",
+        "current_step",
+        "assigned_to",
+        "completed_at"
+      ],
+      "expected_status": "completed",
+      "require_completed_at": true
+    },
+    "activity-log": {
+      "expected_action_pattern": "Completed",
+      "freshness_minutes": 30
+    },
+    "completion-checklist": {
+      "require_all_checked": true
+    },
+    "platform-sync": {
+      "when": "issue_number_exists",
+      "expected_comment_marker": "<!-- sync-issue:{task-id}:summary -->",
+      "verify_task_comment_content": true,
+      "sync_checked_requirements": true,
+      "verify_issue_type": true,
+      "verify_issue_fields": false,
+      "verify_milestone": true,
+      "expected_comment_marker_key": "summary"
+    },
+    "artifact": {
+      "file_pattern": "task.md",
+      "required_sections": [
+        "状态核对"
+      ],
+      "required_patterns": [
+        "^\\$ "
+      ],
+      "freshness_minutes": 30
+    }
+  }
+}

package/templates/.agents/skills/create-pr/config/verify.json

@@ -24,6 +24,7 @@
       "verify_in_labels_match_pr": true,
       "verify_pr_type_label": true,
       "verify_pr_assignee": true,
+      "verify_issue_fields": true,
       "verify_milestone": true,
       "expected_pr_comment_marker_key": "prSummary"
     }

package/templates/.agents/skills/create-task/SKILL.en.md

@@ -78,8 +78,8 @@ created_at: {YYYY-MM-DD HH:mm:ss±HH:MM}
 updated_at: {YYYY-MM-DD HH:mm:ss±HH:MM}
 agent_infra_version: {agent_infra_version}
 created_by: human
-priority:                  # optional; Urgent | High | Medium | Low
-effort:                    # optional; High | Medium | Low
+priority:                  # required; inferred by the AI from the title/description; Urgent | High | Medium | Low
+effort:                    # required; inferred by the AI from the title/description; High | Medium | Low
 start_date:                # optional; YYYY-MM-DD
 target_date:               # optional; YYYY-MM-DD
 current_step: requirement-analysis
@@ -87,7 +87,7 @@ assigned_to: {current AI agent}
 ```
 
 Note: `created_by` is `human` because the task comes from the user's description.
-Optional Issue field metadata may be left empty at task creation; do not invent dates.
+priority / effort are required: the AI infers them from the task title and description (candidates in `.agents/rules/issue-fields.md`; normalize localized input). Leave start_date / target_date empty at creation - analyze-task / plan-task fill them later; do not invent dates.
 
 ### 3. Update Task Status
 

package/templates/.agents/skills/create-task/SKILL.zh-CN.md

@@ -78,8 +78,8 @@ created_at: {YYYY-MM-DD HH:mm:ss±HH:MM}
 updated_at: {YYYY-MM-DD HH:mm:ss±HH:MM}
 agent_infra_version: {agent_infra_version}
 created_by: human
-priority:                  # 可选；Urgent | High | Medium | Low
-effort:                    # 可选；High | Medium | Low
+priority:                  # 必填；由 AI 从标题/描述推断；Urgent | High | Medium | Low
+effort:                    # 必填；由 AI 从标题/描述推断；High | Medium | Low
 start_date:                # 可选；YYYY-MM-DD
 target_date:               # 可选；YYYY-MM-DD
 current_step: requirement-analysis
@@ -87,7 +87,7 @@ assigned_to: {当前 AI 代理}
 ```
 
 注意：`created_by` 为 `human`，因为任务来源于用户的描述。
-可选 Issue 字段元数据在创建任务时可留空；不要臆测日期。
+priority / effort 必填：由 AI 从任务标题与描述推断后填入（候选值见 `.agents/rules/issue-fields.md`；中文输入按本地化映射规范化）。start_date / target_date 创建时保持留空，由 analyze-task / plan-task 阶段填入；不要臆测日期。
 
 ### 3. 更新任务状态
 

package/templates/.agents/skills/implement-task/SKILL.en.md

@@ -27,6 +27,20 @@ Before implementing, stop if any of these thoughts appear. They are violations:
 | "This plan is flawed, so I will improve it as I go." | Deviations from `{plan-artifact}` must be recorded in the report; if the direction is questionable, stop and confirm instead of changing it silently. |
 | "The tests passed, so I can commit this too." | This skill never runs `git add` or `git commit`; committing is a separate step explicitly requested by the user. |
 
+## Step 0: State Check (pre-execution hard gate)
+
+After loading workflow / skill / rules instructions, and before any task-state judgment or user-visible conclusion, run the state check first. Reading instruction files does not count as an external-state action or conclusion.
+
+Run these commands and paste the raw output into both the user-facing reply and this round's `## State Check` section:
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+Before the state check is complete, do not make external-state assertions such as "the code is unchanged", "tests passed", or "there are no other references", including in reasoning. This gate is only a structural floor; evidence pairing and authenticity still require the report template and review discipline.
+
 ## Steps
 
 ### 1. Verify Prerequisites

package/templates/.agents/skills/implement-task/SKILL.zh-CN.md

@@ -27,6 +27,20 @@ description: "根据技术方案实施任务并输出报告"
 | 「方案这里不合理，顺手改更好」 | 偏离 `{plan-artifact}` 必须在报告中记录原因；有异议先停下确认，不擅自改方向。 |
 | 「测试过了，顺便提交一下」 | 本技能绝不执行 `git add`/`git commit`，提交是用户显式发起的独立步骤。 |
 
+## 第 0 步：状态核对（执行前硬约束）
+
+在加载 workflow / skill / rules 指令之后、做任何任务状态判断或用户可见结论之前，必须先执行状态核对。指令类文件读取不算对外动作或结论。
+
+运行以下命令，并把原文粘贴到回复正文和本轮产物的 `## 状态核对` 段：
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+状态核对完成前，禁止任何关于外部状态的断言（例如“代码没变”“测试已通过”“没有其他引用”），包括思考阶段。本门禁只提供结构下限；逐条证据配对和真实性仍需按报告模板与审查要求核对。
+
 ## 执行步骤
 
 ### 1. 验证前置条件

package/templates/.agents/skills/implement-task/config/verify.en.json

@@ -0,0 +1,51 @@
+{
+  "skill": "implement-task",
+  "checks": {
+    "task-meta": {
+      "required_fields": [
+        "id",
+        "type",
+        "workflow",
+        "status",
+        "created_at",
+        "updated_at",
+        "agent_infra_version",
+        "current_step",
+        "assigned_to"
+      ],
+      "expected_step": "implementation"
+    },
+    "artifact": {
+      "file_pattern": "implementation.md|implementation-r{N}.md",
+      "required_sections": [
+        "Modified Files",
+        "Key Code Explanation",
+        "Test Results",
+        "Differences from Plan",
+        "Items for Review",
+        "State Check",
+        "Evidence"
+      ],
+      "freshness_minutes": 30,
+      "required_patterns": [
+        "^\\$ "
+      ]
+    },
+    "activity-log": {
+      "expected_action_pattern": "Implementation \\(Round \\d+\\)",
+      "freshness_minutes": 30
+    },
+    "platform-sync": {
+      "when": "issue_number_exists",
+      "expected_status_label": "status: in-progress",
+      "expected_comment_marker": "<!-- sync-issue:{task-id}:{artifact-stem} -->",
+      "verify_comment_content": true,
+      "verify_task_comment_content": true,
+      "verify_issue_type": true,
+      "verify_issue_fields": false,
+      "verify_milestone": true,
+      "expected_status_label_key": "inProgress",
+      "expected_comment_marker_key": "artifact"
+    }
+  }
+}

package/templates/.agents/skills/implement-task/config/{verify.json → verify.zh-CN.json}

@@ -22,9 +22,14 @@
         "关键代码说明",
         "测试结果",
         "与方案的差异",
-        "供审查关注的内容"
+        "供审查关注的内容",
+        "状态核对",
+        "证据原文"
       ],
-      "freshness_minutes": 30
+      "freshness_minutes": 30,
+      "required_patterns": [
+        "^\\$ "
+      ]
     },
     "activity-log": {
       "expected_action_pattern": "Implementation \\(Round \\d+\\)",

package/templates/.agents/skills/implement-task/reference/report-template.en.md

@@ -10,6 +10,10 @@ Use this structure when creating `implementation.md` or `implementation-r{N}.md`
 - **Implementation Round**: Round {implementation-round}
 - **Artifact File**: `{implementation-artifact}`
 
+## State Check
+
+> Paste the raw state-check command output; each command starts with `$ `.
+
 ## Modified Files
 
 ### New Files
@@ -43,6 +47,17 @@ Use this structure when creating `implementation.md` or `implementation-r{N}.md`
 {test-run-output}
 ```
 
+
+## Evidence
+
+> Pair each "I verified X" claim with the corresponding raw tool output; the gate only checks that this section exists and at least one `$ ` line is present.
+
+- Claim: {verified claim}
+```text
+$ {command}
+{raw output}
+```
+
 ## Differences from Plan
 
 {describe any deviation from the approved plan}

package/templates/.agents/skills/implement-task/reference/report-template.zh-CN.md

@@ -10,6 +10,10 @@
 - **实现轮次**: Round {implementation-round}
 - **产物文件**: `{implementation-artifact}`
 
+## 状态核对
+
+> 粘贴状态核对命令原文；每条命令以 `$ ` 开头。
+
 ## 变更文件
 
 ### 新建文件
@@ -43,6 +47,17 @@
 {test-run-output}
 ```
 
+
+## 证据原文
+
+> 每条“我验证了 X”断言都要配对对应 tool output 原文；gate 仅校验本段存在和至少一行 `$ `。
+
+- 断言：{verified claim}
+```text
+$ {command}
+{raw output}
+```
+
 ## 与方案的差异
 
 {describe any deviation from the approved plan}

package/templates/.agents/skills/plan-task/SKILL.en.md

@@ -13,6 +13,20 @@ description: "Design a technical plan for a task"
 
 Version stamp rule: when creating or updating `task.md` frontmatter, read `.agents/rules/version-stamp.md` first and write or refresh `agent_infra_version`.
 
+## Step 0: State Check (pre-execution hard gate)
+
+After loading workflow / skill / rules instructions, and before any task-state judgment or user-visible conclusion, run the state check first. Reading instruction files does not count as an external-state action or conclusion.
+
+Run these commands and paste the raw output into both the user-facing reply and this round's `## State Check` section:
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+Before the state check is complete, do not make external-state assertions such as "the code is unchanged", "tests passed", or "there are no other references", including in reasoning. This gate is only a structural floor; evidence pairing and authenticity still require the report template and review discipline.
+
 ## Steps
 
 ### 1. Verify Prerequisites
@@ -91,9 +105,18 @@ Update `.agents/workspace/active/{task-id}/task.md`:
 - `assigned_to`: {current AI agent}
 - `updated_at`: {current time}
 - `agent_infra_version`: value from `.agents/rules/version-stamp.md`
+- If `target_date` is empty, write an estimated completion date based on the effort estimate (`YYYY-MM-DD`); leave it empty without blocking when no reasonable estimate exists; keep any existing value
 - Record the plan artifact for this round: `{plan-artifact}` (Round `{plan-round}`)
 - If the task template contains a `## Design` section, update it to link to `{plan-artifact}`
 - Mark technical-design as complete in workflow progress and include the actual round when the task template supports it
+- Before appending the workflow Activity Log entry, re-estimate `effort` based on the technical plan (number of implementation steps, files touched, test matrix scope, integration surface). If the re-estimated value differs from the current value in `task.md`:
+  - Overwrite the `effort` field in frontmatter with the new value
+  - Prepend an Activity Log entry recording the transition (placed before the `Technical Design (Round N)` entry):
+    ```
+    - {YYYY-MM-DD HH:mm:ss±HH:MM} — **Plan Re-estimate** by {agent} — effort {old} → {new} (rationale: {short basis grounded in this plan})
+    ```
+  Both entries may share the same timestamp; ordering is conveyed by list position only.
+  If the re-estimated value matches the current value, skip the Re-estimate entry. The Flow A sync that follows reads the possibly updated frontmatter and propagates the new value to the Issue automatically.
 - **Append** to `## Activity Log` (do NOT overwrite previous entries):
   ```
   - {YYYY-MM-DD HH:mm:ss±HH:MM} — **Technical Design (Round {N})** by {agent} — Plan completed, awaiting human review → {artifact-filename}
@@ -104,6 +127,7 @@ If task.md contains a valid `issue_number`, perform these sync actions (skip and
 - Set `status: pending-design-work` by following issue-sync.md
 - Create or update the task comment marker defined in `.agents/rules/issue-sync.md` (follow the task.md comment sync rule in issue-sync.md)
 - Publish the `{plan-artifact}` comment
+- Read `.agents/rules/issue-fields.md` and follow Flow A to sync every non-empty Issue field (`priority`/`effort`/`start_date`/`target_date`) from `task.md` to the Issue (idempotent; skip without blocking when `has_push=false` or the fetch/write fails)
 
 ### 8. Verification Gate
 

package/templates/.agents/skills/plan-task/SKILL.zh-CN.md

@@ -13,6 +13,20 @@ description: "为任务设计技术方案和实施计划"
 
 版本戳规则：创建或更新 `task.md` frontmatter 时，先读取 `.agents/rules/version-stamp.md`，并写入或刷新 `agent_infra_version`。
 
+## 第 0 步：状态核对（执行前硬约束）
+
+在加载 workflow / skill / rules 指令之后、做任何任务状态判断或用户可见结论之前，必须先执行状态核对。指令类文件读取不算对外动作或结论。
+
+运行以下命令，并把原文粘贴到回复正文和本轮产物的 `## 状态核对` 段：
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+状态核对完成前，禁止任何关于外部状态的断言（例如“代码没变”“测试已通过”“没有其他引用”），包括思考阶段。本门禁只提供结构下限；逐条证据配对和真实性仍需按报告模板与审查要求核对。
+
 ## 执行步骤
 
 ### 1. 验证前置条件
@@ -91,9 +105,18 @@ date "+%Y-%m-%d %H:%M:%S%:z"
 - `assigned_to`：{当前 AI 代理}
 - `updated_at`：{当前时间}
 - `agent_infra_version`：按 `.agents/rules/version-stamp.md` 取值
+- 若 `target_date` 为空，基于工作量评估写入预估完成日（`YYYY-MM-DD`）；无法合理预估时保持留空、不阻塞；已有值则保留
 - 记录本轮方案产物：`{plan-artifact}`（Round `{plan-round}`）
 - 如任务模板包含 `## 设计` 段落，更新为指向 `{plan-artifact}` 的链接
 - 在工作流进度中标记 technical-design 为已完成，并注明实际轮次（如果任务模板支持）
+- 在追加工作流 Activity Log 条目之前，基于技术方案（实施步骤数、涉及文件、测试矩阵范围、集成面）重估 `effort`。若重估值与 `task.md` 当前值不一致：
+  - 用新值覆盖 frontmatter 的 `effort` 字段
+  - 在 `Technical Design (Round N)` 条目之前追加一条转移记录：
+    ```
+    - {YYYY-MM-DD HH:mm:ss±HH:MM} — **Plan Re-estimate** by {agent} — effort {old} → {new} (rationale: {基于本轮方案的简短依据})
+    ```
+  两条条目可共用同一时间戳，顺序仅通过列表位置表达。
+  若重估值与当前值一致，跳过 Re-estimate 条目。后续 Flow A 同步会读取可能更新过的 frontmatter，并自动把新值同步到 Issue。
 - **追加**到 `## Activity Log`（不要覆盖之前的记录）：
   ```
   - {YYYY-MM-DD HH:mm:ss±HH:MM} — **Technical Design (Round {N})** by {agent} — Plan completed, awaiting human review → {artifact-filename}
@@ -104,6 +127,7 @@ date "+%Y-%m-%d %H:%M:%S%:z"
 - 按 issue-sync.md 设置 `status: pending-design-work`
 - 创建或更新 `.agents/rules/issue-sync.md` 中定义的 task 评论标记（按 issue-sync.md 的 task.md 评论同步规则）
 - 发布 `{plan-artifact}` 评论
+- 读取 `.agents/rules/issue-fields.md`，按流程 A 把 `task.md` 中所有非空的 Issue 字段（`priority`/`effort`/`start_date`/`target_date`）同步到 Issue（幂等；`has_push=false` 或取数/写入失败时跳过，不阻断）
 
 ### 8. 完成校验
 

package/templates/.agents/skills/plan-task/config/verify.en.json

@@ -0,0 +1,52 @@
+{
+  "skill": "plan-task",
+  "checks": {
+    "task-meta": {
+      "required_fields": [
+        "id",
+        "type",
+        "workflow",
+        "status",
+        "created_at",
+        "updated_at",
+        "agent_infra_version",
+        "current_step",
+        "assigned_to"
+      ],
+      "expected_step": "technical-design"
+    },
+    "artifact": {
+      "file_pattern": "plan.md|plan-r{N}.md",
+      "required_sections": [
+        "Problem Understanding",
+        "Constraints",
+        "Option Comparison",
+        "Technical Approach",
+        "Implementation Steps",
+        "File List",
+        "Verification Strategy",
+        "State Check"
+      ],
+      "freshness_minutes": 30,
+      "required_patterns": [
+        "^\\$ "
+      ]
+    },
+    "activity-log": {
+      "expected_action_pattern": "Technical Design \\(Round \\d+\\)",
+      "freshness_minutes": 30
+    },
+    "platform-sync": {
+      "when": "issue_number_exists",
+      "expected_status_label": "status: pending-design-work",
+      "expected_comment_marker": "<!-- sync-issue:{task-id}:{artifact-stem} -->",
+      "verify_comment_content": true,
+      "verify_task_comment_content": true,
+      "verify_issue_type": true,
+      "verify_issue_fields": false,
+      "verify_milestone": true,
+      "expected_status_label_key": "pendingDesignWork",
+      "expected_comment_marker_key": "artifact"
+    }
+  }
+}

package/templates/.agents/skills/plan-task/config/{verify.json → verify.zh-CN.json}

@@ -24,9 +24,13 @@
         "技术方法",
         "实施步骤",
         "文件清单",
-        "验证策略"
+        "验证策略",
+        "状态核对"
       ],
-      "freshness_minutes": 30
+      "freshness_minutes": 30,
+      "required_patterns": [
+        "^\\$ "
+      ]
     },
     "activity-log": {
       "expected_action_pattern": "Technical Design \\(Round \\d+\\)",

package/templates/.agents/skills/post-release/SKILL.en.md

@@ -54,6 +54,7 @@ git status --short
 <!-- TODO: Add project-specific follow-up tasks such as demo capture, docs publishing, or downstream notifications -->
 
 - Examples: record a terminal demo, refresh a docs site, notify downstream teams, update release dashboards
+- If a task here displays or captures the just-released version (e.g., recording a CLI demo), move it **before** Step 3 — otherwise it will reflect the next development version, not the release
 - If there are no extra tasks, remove this step in the project-specific copy
 
 ### 6. Create the Follow-up Commit

package/templates/.agents/skills/post-release/SKILL.zh-CN.md

@@ -54,6 +54,7 @@ git status --short
 <!-- TODO: 添加项目特定的发布后任务，例如录制演示、发布文档站或通知下游 -->
 
 - 示例：录制终端演示、刷新文档站、通知下游团队、更新发布面板
+- 如果这一步里的任务需要显示或捕获刚发布的版本号（例如录制 CLI 演示），把它挪到**第 3 步「准备下一个开发版本」之前**——否则会反映下一个开发版本号，而不是 released 版本号
 - 如果没有额外任务，请在项目特化版本中删除此步骤
 
 ### 6. 创建后处理提交

package/templates/.agents/skills/refine-task/SKILL.en.md

@@ -24,6 +24,20 @@ Version stamp rule: when creating or updating `task.md` frontmatter, read `.agen
 | "I can clean up nearby code while I am here." | Scope is limited to review findings; changes outside that scope are out of bounds. |
 | "Clearing env-blocked items makes the report look better." | env-blocked findings are outside the repair scope; preserve and label them truthfully, and never use no-op edits to pad the fix count. |
 
+## Step 0: State Check (pre-execution hard gate)
+
+After loading workflow / skill / rules instructions, and before any task-state judgment or user-visible conclusion, run the state check first. Reading instruction files does not count as an external-state action or conclusion.
+
+Run these commands and paste the raw output into both the user-facing reply and this round's `## State Check` section:
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+Before the state check is complete, do not make external-state assertions such as "the code is unchanged", "tests passed", or "there are no other references", including in reasoning. This gate is only a structural floor; evidence pairing and authenticity still require the report template and review discipline.
+
 ## Steps
 
 ### 1. Verify Prerequisites

package/templates/.agents/skills/refine-task/SKILL.zh-CN.md

@@ -24,6 +24,20 @@ description: "处理代码审查反馈并修复问题"
 | 「顺手把旁边也优化了」 | 范围严格限定在审查项内，范围外改动属于越界。 |
 | 「env-blocked 项清掉报告好看」 | env-blocked 不在修复范围，必须如实保留并标注，禁止 no-op 改动充数。 |
 
+## 第 0 步：状态核对（执行前硬约束）
+
+在加载 workflow / skill / rules 指令之后、做任何任务状态判断或用户可见结论之前，必须先执行状态核对。指令类文件读取不算对外动作或结论。
+
+运行以下命令，并把原文粘贴到回复正文和本轮产物的 `## 状态核对` 段：
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+状态核对完成前，禁止任何关于外部状态的断言（例如“代码没变”“测试已通过”“没有其他引用”），包括思考阶段。本门禁只提供结构下限；逐条证据配对和真实性仍需按报告模板与审查要求核对。
+
 ## 执行步骤
 
 ### 1. 验证前置条件

package/templates/.agents/skills/refine-task/config/verify.en.json

@@ -0,0 +1,47 @@
+{
+  "skill": "refine-task",
+  "checks": {
+    "task-meta": {
+      "required_fields": [
+        "id",
+        "type",
+        "workflow",
+        "status",
+        "created_at",
+        "updated_at",
+        "agent_infra_version",
+        "current_step",
+        "assigned_to"
+      ]
+    },
+    "artifact": {
+      "file_pattern": "refinement.md|refinement-r{N}.md",
+      "required_sections": [
+        "Review Feedback Handling",
+        "Test Results After Refinement",
+        "State Check",
+        "Evidence"
+      ],
+      "freshness_minutes": 30,
+      "required_patterns": [
+        "^\\$ "
+      ]
+    },
+    "activity-log": {
+      "expected_action_pattern": "Refinement \\(Round \\d+, for .+\\)",
+      "freshness_minutes": 30
+    },
+    "platform-sync": {
+      "when": "issue_number_exists",
+      "expected_status_label": "status: in-progress",
+      "expected_comment_marker": "<!-- sync-issue:{task-id}:{artifact-stem} -->",
+      "verify_comment_content": true,
+      "verify_task_comment_content": true,
+      "verify_issue_type": true,
+      "verify_issue_fields": false,
+      "verify_milestone": true,
+      "expected_status_label_key": "inProgress",
+      "expected_comment_marker_key": "artifact"
+    }
+  }
+}

package/templates/.agents/skills/refine-task/config/{verify.json → verify.zh-CN.json}

@@ -18,9 +18,14 @@
       "file_pattern": "refinement.md|refinement-r{N}.md",
       "required_sections": [
         "审查反馈处理",
-        "修复后的测试结果"
+        "修复后的测试结果",
+        "状态核对",
+        "证据原文"
       ],
-      "freshness_minutes": 30
+      "freshness_minutes": 30,
+      "required_patterns": [
+        "^\\$ "
+      ]
     },
     "activity-log": {
       "expected_action_pattern": "Refinement \\(Round \\d+, for .+\\)",

package/templates/.agents/skills/refine-task/reference/report-template.en.md

@@ -12,6 +12,10 @@ Use this structure when writing `refinement.md` or `refinement-r{N}.md`.
 - **Review Input**: `{review-artifact}`
 - **Implementation Context**: `{implementation-artifact}`
 
+## State Check
+
+> Paste the raw state-check command output; each command starts with `$ `.
+
 ### Review Feedback Handling
 
 #### Blocker Fixes
@@ -43,6 +47,17 @@ Use this structure when writing `refinement.md` or `refinement-r{N}.md`.
 #### Unresolved Issues
 - {issue}: {reason}
 
+
+## Evidence
+
+> Pair each "I verified X" claim with the corresponding raw tool output; the gate only checks that this section exists and at least one `$ ` line is present.
+
+- Claim: {verified claim}
+```text
+$ {command}
+{raw output}
+```
+
 ### Test Results After Refinement
 - All tests passing: {yes/no}
 - Test output: {summary}

package/templates/.agents/skills/refine-task/reference/report-template.zh-CN.md

@@ -12,6 +12,10 @@
 - **审查输入**: `{review-artifact}`
 - **实现上下文**: `{implementation-artifact}`
 
+## 状态核对
+
+> 粘贴状态核对命令原文；每条命令以 `$ ` 开头。
+
 ### 审查反馈处理
 
 #### 阻塞项修复
@@ -43,6 +47,17 @@
 #### 未解决问题
 - {issue}: {reason}
 
+
+## 证据原文
+
+> 每条“我验证了 X”断言都要配对对应 tool output 原文；gate 仅校验本段存在和至少一行 `$ `。
+
+- 断言：{verified claim}
+```text
+$ {command}
+{raw output}
+```
+
 ### 修复后的测试结果
 - 所有测试通过: {yes/no}
 - 测试输出: {summary}

package/templates/.agents/skills/review-task/SKILL.en.md

@@ -22,6 +22,20 @@ Version stamp rule: when creating or updating `task.md` frontmatter, read `.agen
 | "It looks mostly fine, so approve it." | The verdict must be backed by blocker/major/minor counts, and every finding must cite file:line; do not approve from impression. |
 | "The test change looks reasonable, so I can skim it." | Before reviewing test changes, check `.agents/rules/testing-discipline.md` item by item (see the step 4 gate). |
 
+## Step 0: State Check (pre-execution hard gate)
+
+After loading workflow / skill / rules instructions, and before any task-state judgment or user-visible conclusion, run the state check first. Reading instruction files does not count as an external-state action or conclusion.
+
+Run these commands and paste the raw output into both the user-facing reply and this round's `## State Check` section:
+
+```bash
+git status -s
+ls -la .agents/workspace/active/{task-id}/
+tail .agents/workspace/active/{task-id}/task.md
+```
+
+Before the state check is complete, do not make external-state assertions such as "the code is unchanged", "tests passed", or "there are no other references", including in reasoning. This gate is only a structural floor; evidence pairing and authenticity still require the report template and review discipline.
+
 ## Steps
 
 ### 1. Verify Prerequisites