OpenClaw

task-finish-contract

Enforce task completion with explicit Goal/Progress/Next state. Prevent mid-task stalls and ensure evidence-based completion.

dalomeve0 来源安装数☆ 0 仓库 Stars

查看来源 ↗安装与使用 SKILL.md ↗

安装与使用

请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求，再添加到工作流。

SkillHub CLI

skillhub install dalomeve/task-finish-contract

打开来源仓库，查看说明和兼容环境后，按来源提供的命令或步骤安装。

将来源链接提供给你的 Agent，让它先读取说明、核对依赖，并向你说明将执行的安装步骤。

查看来源 ↗使用指南

技能说明

Task Finish Contract

Prevent mid-task stalls. Every task must finish with explicit state and evidence.

Problem

Agents often:

Stop mid-task without explanation
Output plans without execution
Lack clear completion criteria
Missing evidence artifacts

Workflow

1. State Output (Each Substantial Step)

**Goal**: What finished looks like
**Progress**: What has been done
**Next**: One concrete action executing now

2. Completion Proof Format

For tasks with 2+ steps, include:

**DONE_CHECKLIST**:
- [ ] Item 1 completed
- [ ] Item 2 completed

**EVIDENCE**:
- Executed: command/action summary
- Artifact: path/URL/id
- Verified: check command result

**NEXT_AUTONOMOUS_STEP**:
- One follow-up that runs without user input

3. Anti-Stall Rule

Planning-only replies: max 1
Next reply MUST contain execution evidence
Never end with "I will now..." without tool result

Executable Completion Criteria

Criteria	Verification
Goal stated	`Select-String "Goal" memory/{date}.md` matches
Progress tracked	`Select-String "Progress" memory/{date}.md` matches
Next action defined	`Select-String "Next" memory/{date}.md` matches
Evidence present	Artifact path/URL exists
No unresolved markers	`Select-String "TODO

Privacy/Safety

No sensitive data in completion evidence
Artifact paths use relative or workspace paths
No credentials in task logs

Self-Use Trigger

Use when:

Starting any multi-step task
Resuming after interruption
Handoff to another agent

Finish what you start. Prove it with evidence.

来源信息

Source: github
Repository: openclaw/skills
Path: skills/dalomeve/task-finish-contract
Version: —
Commit: 24fa75078ba3cd257ba6c16d2f41e309fed09c94
分类来源: rules
状态: active
可见性: public

这些信息来自收录时的来源分析，可能滞后；使用前请核对最新来源文件。

来源内容

SKILL.md

# Task Finish Contract

Prevent mid-task stalls. Every task must finish with explicit state and evidence.

## Problem

Agents often:
- Stop mid-task without explanation
- Output plans without execution
- Lack clear completion criteria
- Missing evidence artifacts

## Workflow

### 1. State Output (Each Substantial Step)

```markdown
**Goal**: What finished looks like
**Progress**: What has been done
**Next**: One concrete action executing now
```

### 2. Completion Proof Format

For tasks with 2+ steps, include:

```markdown
**DONE_CHECKLIST**:
- [ ] Item 1 completed
- [ ] Item 2 completed

**EVIDENCE**:
- Executed: command/action summary
- Artifact: path/URL/id
- Verified: check command result

**NEXT_AUTONOMOUS_STEP**:
- One follow-up that runs without user input
```

### 3. Anti-Stall Rule

- Planning-only replies: max 1
- Next reply MUST contain execution evidence
- Never end with "I will now..." without tool result

## Executable Completion Criteria

| Criteria | Verification |
|----------|-------------|
| Goal stated | `Select-String "Goal" memory/{date}.md` matches |
| Progress tracked | `Select-String "Progress" memory/{date}.md` matches |
| Next action defined | `Select-String "Next" memory/{date}.md` matches |
| Evidence present | Artifact path/URL exists |
| No unresolved markers | `Select-String "TODO|PENDING|TBD" artifact` returns nothing |

## Privacy/Safety

- No sensitive data in completion evidence
- Artifact paths use relative or workspace paths
- No credentials in task logs

## Self-Use Trigger

Use when:
- Starting any multi-step task
- Resuming after interruption
- Handoff to another agent

---

**Finish what you start. Prove it with evidence.**

搜索