top-prompt-cleaner

分类: 工具与效率 | 上传者: IvanDembicki | 下载: 0 | 版本: v1.0（最新）

清理、构建、验证和调整任何单个提示。将复杂的多步骤工作流程升级为 TOP Skill Factory。

更新日志: Source: GitHub https://github.com/IvanDembicki/TOP

目录结构

当前层级: tree/main/top-prompt-cleaner/

📁 .github/
- 📁 workflows/
  - 📄 validate.yml 424 B
📁 docs/
- 📁 migration/
  - 📄 frontend_adapter.md 587 B
  - 📄 frontend_style.md 1.0 KB
- 📄 ai_usage.md 145 B
- 📄 contracts.md 8.8 KB
- 📄 security.md 2.3 KB
- 📄 troubleshooting.md 4.6 KB
- 📄 usage.md 4.0 KB
📁 product/
- 📄 landing.md 3.8 KB
- 📄 objections.md 227 B
- 📄 onboarding.md 168 B
- 📄 output-layout.md 3.9 KB
- 📄 user-flow.md 3.3 KB
- 📄 value.md 108 B
📁 scripts/
- 📄 check-updates.js 2.7 KB
- 📄 validate.js 11.1 KB
📁 top/
- 📁 examples/
  - 📄 example-blocking-conflict.md 4.1 KB
  - 📄 example-clarification-with-u.md 5.9 KB
  - 📄 example-escalation.md 2.0 KB
  - 📄 example-generic-style.md 4.5 KB
  - 📄 example-gpt-style.md 5.7 KB
  - 📄 example-medium-complexity.md 4.9 KB
  - 📄 example-quick-clean.md 3.7 KB
  - 📄 example-sensitive-blocked.md 2.4 KB
  - 📄 example-strict-blocked.md 4.0 KB
  - 📄 example-target-style.md 8.4 KB
  - 📄 README.md 2.8 KB
- 📁 model-profiles/
  - 📄 claude.md 2.2 KB
  - 📄 custom.md 2.4 KB
  - 📄 generic.md 2.4 KB
  - 📄 gpt.md 2.7 KB
- 📁 modes/
  - 📄 quick_clean.md 1.6 KB
  - 📄 strict.md 1.4 KB
  - 📄 target_llm_style.md 2.9 KB
- 📁 prompts/
  - 📁 nodes/
    
    📄 clarification-controller.md 3.8 KB
    
    📄 complexity-detector.md 2.0 KB
    
    📄 conflict-detector.md 1.5 KB
    
    📄 sensitive-data-detector.md 3.6 KB
    
    📄 structure-extractor.md 1.2 KB
  - 📄 factory_escalation.md 1.1 KB
  - 📄 final-decision-controller.md 1.2 KB
  - 📄 input-controller.md 1.5 KB
  - 📄 mode-router.md 1.9 KB
  - 📄 output-builder.md 1.5 KB
  - 📄 prompt_analyzer.md 1.2 KB
  - 📄 validation-controller.md 1.8 KB
- 📁 roadmap/
  - 📄 batch_clean.md 2.2 KB
- 📁 schemas/
  - 📄 clarification_request.schema.json 2.3 KB
  - 📄 clarification_state.schema.json 1.7 KB
  - 📄 complexity_report.schema.json 1.5 KB
  - 📄 conflict_report.schema.json 1.9 KB
  - 📄 diff.schema.json 2.8 KB
  - 📄 extraction_result.schema.json 1.7 KB
  - 📄 final_decision_signal.schema.json 1.3 KB
  - 📄 final_output.schema.json 3.8 KB
  - 📄 mode_routing_result.schema.json 1.3 KB
  - 📄 normalized_input.schema.json 1.6 KB
  - 📄 release-metadata.schema.json 943 B
  - 📄 sensitive_data_report.schema.json 2.1 KB
  - 📄 structured_prompt.json 2.4 KB
  - 📄 target_style_output.schema.json 3.0 KB
  - 📄 user_response.schema.json 845 B
  - 📄 validation_result.schema.json 1.8 KB
- 📁 shared-rules/
  - 📄 core-invariants.md 1.3 KB
  - 📄 startup-update-check.md 1000 B
- 📁 validation/
  - 📄 rules.md 1.4 KB
- 📄 spec.json 6.0 KB
📄 .gitignore 40 B
📄 LICENSE 1.1 KB
📄 meta-prompt.md 4.4 KB
📄 package-lock.json 2.9 KB
📄 package.json 620 B
📄 README.md 4.1 KB
📄 release-metadata.json 470 B
📄 RELEASE_CRITERIA.md 2.9 KB
📄 RELEASE_NOTES.md 2.9 KB
📄 SKILL.md 2.9 KB
📄 VALIDATION_REPORT.md 3.0 KB

SKILL.md

---
name: top-prompt-cleaner
version: 1.0.1
description: Clean, structure, validate, and adapt any single prompt. Escalates complex multi-step workflows to TOP Skill Factory.
trigger: when a user shares a prompt they want improved, cleaned, structured, or adapted for a specific model
---

# TOP Prompt Cleaner

## Startup sequence

On every invocation, before processing the user's prompt:

1. Read `release-metadata.json`.
2. If a trusted comparison manifest is available, compare versions and surface the result.
3. If the user references a different version, or update state is uncertain, re-read `SKILL.md` and `top/spec.json` before proceeding.
4. Proceed to `InputController`.

See `top/shared-rules/startup-update-check.md` for full behavior.

Within the TOP product line, this is the single-prompt cleanup tool. It should escalate to `top-skill-factory` rather than simulate skill design or workflow governance.

---

Use this skill when:
- User gives one prompt and wants it improved, shortened, or restructured
- User asks why their prompt produces inconsistent results
- User wants a prompt adapted for Claude or GPT conventions
- User wants to see what changed and why (diff)

Do not use this skill when:
- Task describes a multi-agent or multi-step AI workflow
- Task requires designing a skill, agent, or system
- Request involves routing between multiple AI behaviors based on runtime conditions
-> For those cases: use TOP Skill Factory in CreateNewSkillMode

## What the user receives

| Output | Description |
|---|---|
| `cleaned_prompt` | Noise-free, contradiction-resolved version |
| `structured_prompt` | Explicit goal / constraints / output_format |
| `diff` | What changed and why |
| `escalation_notice` | When the prompt is actually a workflow |

## Modes

- **QuickCleanMode** - fast single-pass, no questions asked
- **StrictMode** - validates required fields, blocks when goal or output format are missing
- **TargetLLMStyleMode** - adapts style for Claude or GPT conventions

## Key behavior

- High complexity -> escalation to TOP Skill Factory, no partial cleanup
- Sensitive content (API keys, passwords) -> blocked immediately
- PII / internal URLs -> warning in QuickClean, blocked in StrictClean
- Blocking conflicts -> surfaced in diff, not silently resolved
- Missing goal or output_format in StrictMode -> clarification request

## Reading order for reviewers

1. `release-metadata.json` - current release metadata and startup update check policy
2. `top/spec.json` - tree structure, invariants, output contract
3. `top/shared-rules/core-invariants.md` - 16 rules that govern every output
4. `top/shared-rules/startup-update-check.md` - startup update behavior
5. `top/prompts/mode-router.md` - routing logic
6. `top/modes/` - relevant mode
7. `top/schemas/` - all machine-checkable contracts
8. `top/examples/` - worked examples covering all terminal states
9. `RELEASE_NOTES.md` - release history

登录后下载/点赞/收藏 ❤ 9 | ★ 0

top-prompt-cleaner

目录结构

SKILL.md

举报内容

提示