sys-executing-threads

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Thread Executor

线程执行器

Write agent outputs to thread stage files using the numbered convention.

按照编号约定将Agent输出写入线程阶段文件。

Responsibility

职责

Accept agent outputs and write them to the correct numbered stage files in a thread directory. Update frontmatter to track completion.

接收Agent输出，并将其写入线程目录中对应的带编号阶段文件。更新前置元数据以跟踪完成状态。

Thread File Convention

线程文件约定

threads/{domain}/{thread-name}/
├── 1-input.md          # Goal context (written by sys-activating-goals, NEVER by this skill)
├── 2-hypothesis.md     # Approach, assumptions, expected outcomes
├── 3-implication.md    # If hypothesis true, what follows
├── 4-decision.md       # Chosen course of action
├── 5-actions.md        # Execution steps and status
└── 6-learning.md       # Outcomes, insights, improvements

threads/{domain}/{thread-name}/
├── 1-input.md          # 目标上下文（由sys-activating-goals写入，此技能绝不会写入）
├── 2-hypothesis.md     # 方法、假设、预期成果
├── 3-implication.md    # 若假设成立，后续影响
├── 4-decision.md       # 选定的行动方案
├── 5-actions.md        # 执行步骤与状态
└── 6-learning.md       # 成果、洞察、改进建议

Agent-to-Stage Mapping

Agent与阶段映射

Type	Stages Written	Agents
planning	2, 3, 4	mkt-strategist, sls-strategist
execution	5, 6	mkt-campaign-manager, sls-outbound-manager, cst-advocacy-manager
spec	2, 3, 4, 5	prd-engineer, prd-growth-engineer
full-cycle	2, 3, 4, 5, 6	ops-manager
lifecycle	2, 4, 5, 6	cst-success-manager, cst-expansion-manager, cst-retention-manager
content	2, 5	mkt-content-manager, sls-enablement-manager
inbound	2, 3, 5	mkt-inbound-manager
partnership	2, 4, 5, 6	sls-partner-manager

类型	写入阶段	Agent
规划型	2, 3, 4	mkt-strategist, sls-strategist
执行型	5, 6	mkt-campaign-manager, sls-outbound-manager, cst-advocacy-manager
规格型	2, 3, 4, 5	prd-engineer, prd-growth-engineer
全周期型	2, 3, 4, 5, 6	ops-manager
生命周期型	2, 4, 5, 6	cst-success-manager, cst-expansion-manager, cst-retention-manager
内容型	2, 5	mkt-content-manager, sls-enablement-manager
入站型	2, 3, 5	mkt-inbound-manager
合作伙伴型	2, 4, 5, 6	sls-partner-manager

Input

输入

The calling agent provides:

yaml

thread_path: threads/{domain}/{thread-name}
agent_name: {agent that produced the output}
agent_type: planning | execution | spec | full-cycle | lifecycle | content | inbound | partnership
outputs:
  hypothesis: |    # Stage 2 content (if in agent's stages)
    ...
  implication: |   # Stage 3 content
    ...
  decision: |      # Stage 4 content
    ...
  actions: |       # Stage 5 content
    ...
  learning: |      # Stage 6 content
    ...

调用此技能的Agent需提供：

yaml

thread_path: threads/{domain}/{thread-name}
agent_name: {agent that produced the output}
agent_type: planning | execution | spec | full-cycle | lifecycle | content | inbound | partnership
outputs:
  hypothesis: |    # Stage 2 content (if in agent's stages)
    ...
  implication: |   # Stage 3 content
    ...
  decision: |      # Stage 4 content
    ...
  actions: |       # Stage 5 content
    ...
  learning: |      # Stage 6 content
    ...

Process

流程

Step 1: Validate Thread

步骤1：验证线程

1. Confirm thread_path exists
2. Confirm 1-input.md exists (thread was properly activated)
3. Read 1-input.md frontmatter for thread_id and goal_id

1. 确认thread_path存在
2. 确认1-input.md存在（线程已正确激活）
3. 读取1-input.md的前置元数据以获取thread_id和goal_id

Step 2: Determine Stages

步骤2：确定阶段

1. Look up agent_type in mapping table
2. Get list of stage numbers to write
3. Confirm agent provided content for each stage

1. 在映射表中查找agent_type
2. 获取需写入的阶段编号列表
3. 确认Agent已提供每个阶段的内容

Step 3: Write Stage Files

步骤3：写入阶段文件

For each stage in the agent's mapping:

1. Create stage file with frontmatter:
   - status: completed
   - completed_by: {agent_name}
   - completed_at: {YYYY-MM-DD}
2. Write agent output as file body
3. Save file

针对Agent映射中的每个阶段：

1. 创建包含前置元数据的阶段文件：
   - status: completed
   - completed_by: {agent_name}
   - completed_at: {YYYY-MM-DD}
2. 将Agent输出写入文件主体
3. 保存文件

Step 4: Verify

步骤4：验证

1. Confirm all mapped stages have status: completed
2. Return summary of stages written

1. 确认所有映射阶段的状态均为completed
2. 返回已写入阶段的摘要

Frontmatter

前置元数据

Stage files are created by this skill (not pre-existing). Each file gets:

yaml

---
status: completed
completed_by: mkt-strategist
completed_at: 2026-02-01
---

For

4-decision.md

specifically:

yaml

---
status: completed
decided_by: mkt-strategist
decided_at: 2026-02-01
---

For

5-actions.md

specifically:

yaml

---
status: completed
started_at: 2026-02-01
completed_at: 2026-02-01
---

此技能会创建阶段文件（而非使用预先存在的文件）。每个文件都会添加：

yaml

---
status: completed
completed_by: mkt-strategist
completed_at: 2026-02-01
---

针对

4-decision.md

的特殊处理：

yaml

---
status: completed
decided_by: mkt-strategist
decided_at: 2026-02-01
---

针对

5-actions.md

的特殊处理：

yaml

---
status: completed
started_at: 2026-02-01
completed_at: 2026-02-01
---

Stage Content Guidelines

阶段内容规范

2-hypothesis.md

Agent fills:

Approach section with strategy/methodology
Expected Outcome with measurable targets
Key Assumptions table
Risks table

Agent需填写：

方法部分：包含策略/方法论
预期成果：可衡量的目标
关键假设表格
风险表格

3-implication.md

Agent fills:

If Hypothesis Succeeds with quantified impact
If Hypothesis Fails with fallback plan
Decision Criteria table with thresholds

Agent需填写：

若假设成立：量化影响
若假设不成立：备选方案
决策标准表格：包含阈值

4-decision.md

Agent fills:

Chosen Action with explicit commitment
Rationale linking back to hypothesis
Trade-offs table
Approval gate (if impact >= 0.8)

Agent需填写：

选定行动：明确的承诺
理由：关联回假设内容
权衡表格
审批关口（若影响 >= 0.8）

5-actions.md

Agent fills:

Execution Plan table with steps, owners, due dates
Progress Log (updated during execution)
Blockers (if any)

Agent需填写：

执行计划表格：包含步骤、负责人、截止日期
进度日志（执行过程中更新）
阻塞问题（如有）

6-learning.md

Agent fills:

Outcome (expected vs actual)
What Worked list
What Didn't Work list
Key Insights
Process Improvements
Recommendations

Agent需填写：

成果（预期 vs 实际）
有效事项列表
无效事项列表
关键洞察
流程改进建议
后续推荐方案

Output

输出

yaml

execution_summary:
  thread_path: threads/{domain}/{thread-name}
  agent: {agent_name}
  type: {agent_type}
  stages_written: [2, 3, 4]  # example for planning type
  completed_at: 2026-02-01
  status: all_stages_written | partial

yaml

execution_summary:
  thread_path: threads/{domain}/{thread-name}
  agent: {agent_name}
  type: {agent_type}
  stages_written: [2, 3, 4]  # example for planning type
  completed_at: 2026-02-01
  status: all_stages_written | partial

Integration

集成

Upstream

上游

All thread-routable agents: Produce outputs then call this skill
```
sys-activating-goals
```
: Creates thread with 1-input.md (prerequisite)

所有可路由线程的Agent：完成输出后调用此技能
```
sys-activating-goals
```
：创建包含1-input.md的线程（前置条件）

Downstream

下游

```
sys-tracking-goals
```
: Reads completed stages to assess progress
```
6-learning.md
```
: Feeds back into goal tracking and loop detection

```
sys-tracking-goals
```
：读取已完成的阶段以评估进度
```
6-learning.md
```
：反馈到目标跟踪与循环检测

Constraints

约束

This skill NEVER:

Writes to
```
1-input.md
```
(owned by sys-activating-goals)
Creates thread directories (owned by sys-activating-goals)
Executes domain work (agents do)
Skips stages in the agent's mapping

This skill ALWAYS:

Validates thread exists before writing
Updates frontmatter status on every stage file
Records which agent completed each stage
Returns execution summary

此技能绝不会：

写入
```
1-input.md
```
（归sys-activating-goals所有）
创建线程目录（归sys-activating-goals所有）
执行领域工作（由Agent完成）
跳过Agent映射中的阶段

此技能始终会：

在写入前验证线程是否存在
更新每个阶段文件的前置元数据状态
记录完成每个阶段的Agent信息
返回已写入阶段的摘要