SuperAPI AI Directory

harness-auto-full-execution

General & other

View repository

Updated 10-15 17:25

About this skill

harness-auto-full-execution fully automated execution skill

Core Capabilities

  1. Orchestrate all 16 base skills
  2. Execute the complete automated process: initialization → goal → brainstorming → planning → execution → verification → completion
  3. Automatically trigger downstream skills
  4. Monitor execution status
  5. Handle exceptions and circuit breaking
  6. Generate a complete delivery report

Super Combo Skill Description

Skill Type: Super Combo Skill Base Dependencies: All 16 base skills (Phase 1-4) Execution Mode: Orchestrator, does not directly operate on files, triggers other skills through events

Execution Steps

Step 1: Check system status

Use the Read tool to read:.EnjoyHarness/GLOBAL_STATE.md

Check conditions:

  • System is initialized (harness-init completed)
  • No active subagents
  • No tasks in progress

If the system is not initialized:

⚠️ 系统未初始化,自动触发 harness-init...
(触发 harness-init)
等待初始化完成...

Step 1.5: Lock the autonomous execution contract

Use the Read / Edit tools to check and update:.EnjoyHarness/EXECUTION_CONTRACT.md

Requirements:

  • After requirements clarification and brainstorming are complete, set requirements_confirmed to true
  • Record brainstorm_user_participation as the actual mode of this brainstorming session
  • Fix mode as autonomous_after_confirmation
  • The default execution policy remains:
    • execution_mode: parallel_session
    • resume_policy: auto_continue
    • retry_policy: retry_3_then_recover
    • human_escalation_policy: true_blockers_only

If the contract file does not exist, create a default template; if requirements_confirmed: false, only allow continuing with steps related to requirements confirmation, and do not enter the subsequent unattended execution phase. By default,brainstorm_user_participation: false, that is, brainstorming is also completed automatically by the system; only when explicitly configured as true , is the user allowed to participate in brainstorm.

Step 2: Receive user requirements

User requirements format:

需求描述:
{用户原始需求描述}

约束条件(可选):
- 时间约束:{deadline}
- 资源约束:{resource limits}
- 技术约束:{tech constraints}

Step 3: Trigger goal creation

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | SUPER_SKILL_START | harness-auto-full-execution | 接收用户需求 | SUCCESS
{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-goal | 创建SMART目标 | PENDING

Wait for event:

{TIMESTAMP} | SKILL_COMPLETE | harness-goal | 目标创建完成 | SUCCESS

Step 4: Trigger brainstorming

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-brainstorm | 设计讨论 | PENDING

Wait for event:

{TIMESTAMP} | SKILL_COMPLETE | harness-brainstorm | 设计文档完成 | SUCCESS

Note:

  • If brainstorm_user_participation: false, this step is completed fully automatically.
  • If brainstorm_user_participation: true, user participation is only allowed in this step; after final confirmation is completed, the process immediately returns to unattended mode.

Step 5: Trigger plan writing

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-plan | 编写实施计划 | PENDING

Wait for event:

{TIMESTAMP} | SKILL_COMPLETE | harness-plan | 实施计划完成 | SUCCESS

Step 6: Trigger plan execution

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | executing-plans | 执行实施计划 | PENDING

Wait for event:

{TIMESTAMP} | PHASE_COMPLETE | executing-plans | 计划执行完成 | SUCCESS

Execution Monitoring:

  • Periodically read:.EnjoyHarness/GLOBAL_STATE.md(check current_task)
  • Periodically read:.EnjoyHarness/EVENT_LOG.md(check the latest event)
  • Periodically read:.trace/ERROR_TRACE.md(check for errors)
  • Periodically read:.EnjoyHarness/EXECUTION_CONTRACT.md(confirm it is still in autonomous_after_confirmation)

Step 7: Trigger verification

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-validate-output | 验证输出 | PENDING

Wait for event:

{TIMESTAMP} | SKILL_COMPLETE | harness-validate-output | 输出验证通过 | SUCCESS

Step 8: Trigger effect evaluation

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-evaluate-performance | 效果评估 | PENDING

Wait for event:

{TIMESTAMP} | SKILL_COMPLETE | harness-evaluate-performance | 效果评估完成 | SUCCESS

Step 9: Trigger target verification

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-goal | 验证目标达成 | PENDING

Wait for event:

{TIMESTAMP} | SKILL_COMPLETE | harness-goal | 目标验证通过 | SUCCESS

Step 10: Generate delivery report

Use the Write tool to create:.trace/DELIVERY_REPORT_{TIMESTAMP}.md

Report content:

---
created: {TIMESTAMP}
task_id: {TASK_ID}
status: COMPLETE
---

# 交付报告

## 任务概述
{用户需求描述}

## 执行流程
1. ✅ 系统初始化 - harness-init
2. ✅ 目标创建 - harness-goal
3. ✅ 设计讨论 - harness-brainstorm
4. ✅ 计划编写 - harness-plan
5. ✅ 计划执行 - executing-plans
6. ✅ 输出验证 - harness-validate-output
7. ✅ 效果评估 - harness-evaluate-performance
8. ✅ 目标验证 - harness-goal

## 交付物清单
- 设计文档:`docs/plans/{design-doc}.md`
- 实施计划:`docs/plans/{implementation-plan}.md`
- 源代码:{文件列表}
- 测试代码:{测试文件列表}
- 文档:{文档文件列表}

## 质量指标
- 测试覆盖率:{coverage}%
- 代码复杂度:{complexity}
- 性能指标:{performance metrics}
- 目标达成度:{goal achievement}%

## 执行统计
- 总迭代次数:{iterations}
- 总Token消耗:{tokens}
- 总执行时间:{duration}
- 错误次数:{errors}
- 熔断次数:{circuit-breakers}

## 风险与改进
- 识别的风险:{risks}
- 改进建议:{improvements}

## 结论
{任务状态},目标已达成,质量符合预期。

Step 11: Trigger self-evolution

Use the Edit tool to write:.EnjoyHarness/EVENT_LOG.md

Trigger event:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-evolve | 记录经验学习 | PENDING

Step 12: Output completion information

Use the Bash tool to output:

echo ""
echo "✅ harness-auto-full-execution 完成!"
echo ""
echo "📋 交付报告:"
echo " - 文件: .trace/DELIVERY_REPORT_{TIMESTAMP}.md"
echo ""
echo "🎯 交付物:"
echo " - 设计文档: docs/plans/{design-doc}.md"
echo " - 实施计划: docs/plans/{implementation-plan}.md"
echo " - 源代码: {file count}个文件"
echo " - 测试代码: {test count}个文件"
echo ""
echo "📊 质量:"
echo " - 测试覆盖率: {coverage}%"
echo " - 目标达成度: {achievement}%"
echo ""
echo "⏱️ 统计:"
echo " - 执行时间: {duration}"
echo " - Token消耗: {tokens}"
echo " - 迭代次数: {iterations}"
echo ""

Complete automation flow

digraph auto_execution {
    rankdir=TB;

    "用户需求" [shape=box, style=filled, fillcolor="#c8e6c9"];

    subgraph cluster_phase1 {
        label="Phase 1: 初始化";
        style=filled;
        color=lightgrey;

        "检查系统状态" [shape=box];
        "harness-init" [shape=box, style=filled, fillcolor="#bbdefb"];
    }

    subgraph cluster_phase2 {
        label="Phase 2: 设计";
        style=filled;
        color=lightgrey;

        "harness-goal" [shape=box, style=filled, fillcolor="#fff9c4"];
        "harness-brainstorm" [shape=box, style=filled, fillcolor="#fff9c4"];
        "harness-plan" [shape=box, style=filled, fillcolor="#fff9c4"];
    }

    subgraph cluster_phase3 {
        label="Phase 3: 执行";
        style=filled;
        color=lightgrey;

        "executing-plans" [shape=box, style=filled, fillcolor="#f8bbd0"];
        "基础技能编排" [shape=box];
    }

    subgraph cluster_phase4 {
        label="Phase 4: 验证";
        style=filled;
        color=lightgrey;

        "harness-validate-output" [shape=box, style=filled, fillcolor="#e1bee7"];
        "harness-evaluate-performance" [shape=box, style=filled, fillcolor="#e1bee7"];
        "harness-goal验证" [shape=box, style=filled, fillcolor="#e1bee7"];
    }

    subgraph cluster_phase5 {
        label="Phase 5: 完成";
        style=filled;
        color=lightgrey;

        "生成交付报告" [shape=box, style=filled, fillcolor="#81c784"];
        "harness-evolve" [shape=box, style=filled, fillcolor="#81c784"];
        "完成" [shape=doublecircle, style=filled, fillcolor="#81c784"];
    }

    "用户需求" -> "检查系统状态";
    "检查系统状态" -> "harness-init" [label="未初始化"];
    "检查系统状态" -> "harness-goal" [label="已初始化"];
    "harness-init" -> "harness-goal";

    "harness-goal" -> "harness-brainstorm";
    "harness-brainstorm" -> "harness-plan";
    "harness-plan" -> "executing-plans";

    "executing-plans" -> "基础技能编排";
    "基础技能编排" -> "harness-validate-output";

    "harness-validate-output" -> "harness-evaluate-performance";
    "harness-evaluate-performance" -> "harness-goal验证";
    "harness-goal验证" -> "生成交付报告";

    "生成交付报告" -> "harness-evolve";
    "harness-evolve" -> "完成";
}

Exception handling

Exception type 1: Basic skill failure

Detection:

{TIMESTAMP} | SKILL_FAILURE | {skill-name} | {error details} | FAILURE

Handling:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-handle-failure | 处理失败 | PENDING

Wait:

{TIMESTAMP} | SKILL_COMPLETE | harness-handle-failure | 失败已处理 | SUCCESS

Exception type 2: Circuit breaker triggered

Detection:

{TIMESTAMP} | CIRCUIT_BREAKER_TRIGGERED | {type} | {details} | PAUSE

Handling:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-handle-failure | 熔断后自动恢复 | PENDING

Wait:

{TIMESTAMP} | HUMAN_DECISION | {decision} | {details} | RESUME/ABORT

Exception type 3: Target verification failure

Detection:

{TIMESTAMP} | SKILL_FAILURE | harness-goal | 目标未达成 | FAILURE

Handling:

{TIMESTAMP} | LOOP_BACK | executing-plans | 补充缺失部分 | PENDING

Return to Step 6 and continue executing the missing parts.

Exception type 4: Verification failure

Detection:

{TIMESTAMP} | SKILL_FAILURE | harness-validate-output | 验证失败 | FAILURE

Handling:

{TIMESTAMP} | TRIGGER_DOWNSTREAM | harness-diagnose-and-improve | 诊断问题 | PENDING

Fix after diagnosis, then re-verify.

Success criteria

  • .EnjoyHarness/EXECUTION_CONTRACT.md Autonomous execution policy written
  • After requirements confirmation, the user is no longer asked to choose an execution method
  • System initialization check passed
  • Target created successfully
  • Design document generated
  • Implementation plan generated
  • Plan execution completed
  • Output verification passed
  • Effect evaluation completed
  • Target verification passed
  • Delivery report generated
  • Experience recording completed

Failure fallback

  • Basic skill failure → trigger harness-handle-failure
  • Circuit breaker triggered → first trigger harness-handle-failure; escalate to harness-escalate-to-human only for real blockers
  • Target not achieved → return to execution to supplement the missing parts
  • Verification failure → trigger harness-diagnose-and-improve
  • Multiple failures → automatic recovery, automatic splitting, or escalation for real blockers

Linkage relationships

  • Trigger timing: the user raises a requirement and chooses full-process automatic execution
  • Upstream skill: none (super entry point)
  • Downstream skills: all 16 basic skills (triggered through events)
  • Parallel skills: none (serial orchestration)

Key principles

1. Orchestrator pattern

harness-auto-full-execution is an orchestrator; it does not directly manipulate files and only triggers other skills through events:

  • ✅ Correct: write to EVENT_LOG to trigger events and wait for downstream skills to complete
  • ❌ Wrong: directly use Write/Edit tools to modify files

2. Event-driven

All skill coordination is done through EVENT_LOG.md events:

触发:TRIGGER_DOWNSTREAM | {skill} | {action} | PENDING
等待:SKILL_COMPLETE | {skill} | {result} | SUCCESS

3. Monitoring loop

Periodically read GLOBAL_STATE.md and EVENT_LOG.md to monitor execution status:

while not_complete:
    read GLOBAL_STATE.md
    read EVENT_LOG.md
    check current_task
    check latest_events
    sleep 5s

4. Exception propagation

When encountering an exception, do not handle it directly; trigger the corresponding handling skill:

  • Basic skill failure → harness-handle-failure
  • Circuit breaker → harness-escalate-to-human
  • Validation failed → harness-diagnose-and-improve

5. Completeness Guarantee

Ensure the process is executed completely, without omitting any steps:

  • Initialization → Goal → Design → Plan → Execution → Validation → Evaluation → Completion
  • After requirements confirmation, EXECUTION_CONTRACT.md explicitly takes over the unattended strategy
  • Each step has clear success criteria
  • Each step has a clear failure fallback

Usage Example

Example: Add user authentication feature

User input:

需求描述:
为系统添加用户邮箱密码认证功能,支持注册、登录、登出。

约束条件:
- 时间约束:今天完成
- 技术约束:使用bcrypt加密密码,JWT生成token

harness-auto-full-execution execution process:

Phase 1: Initialization

检查系统状态...
✅ 系统已初始化

Phase 2: Design

触发 harness-goal...
✅ 目标创建完成:实现用户邮箱密码认证

触发 harness-brainstorm...
(设计讨论过程)
✅ 设计文档完成:docs/plans/2026-03-28-user-auth-design.md

触发 harness-plan...
✅ 实施计划完成:docs/plans/2026-03-28-user-auth-implementation-plan.md

Phase 3: Execution

触发 executing-plans...
(执行实施计划)
✅ 计划执行完成

Phase 4: Validation

触发 harness-validate-output...
✅ 输出验证通过

触发 harness-evaluate-performance...
✅ 效果评估完成

触发 harness-goal验证...
✅ 目标验证通过

Phase 5: Completion

生成交付报告...
✅ 交付报告:.trace/DELIVERY_REPORT_20260328T170000.md

触发 harness-evolve...
✅ 经验记录完成

✅ 全流程自动执行完成!

Differences from Other Super Combo Skills

Super Combo Skill Goal Features
harness-auto-full-execution Fully automated end-to-end execution Orchestrates all 16 skills, from requirements to delivery
harness-10x-productivity-boost 10x productivity improvement Parallel execution, intelligent scheduling, efficiency optimization

Iteration Count

Estimated number of iterations for this skill execution: about 50-100 times

  • Read status file: 20-30 times
  • Edit EVENT_LOG: 20-30 times
  • Write delivery report: 1 time
  • Bash output: 10-20 times
  • Waiting for downstream skills: not counted

Reference

  • EnjoyHarness Design Document - docs/plans/2026-03-28-enjoyharness-design-v3.md
  • All basic skills in Phase 1-4 - basic orchestration units