360 Secure SkillHub
360 安全技能中心

Promptify Skill

接收用户的提示词,按需派发 clarifier、codebase-researcher、web-researcher 子代理补充上下文,去除冗余措辞并按 Role/Task/Constraints/Output 四要素重写,最后输出结构化提示词、复制命令与简短说明。

tolibear
tolibear
热度 96.12
最近更新 2026-05-17 11:10:01
分类:效率依赖:内部调用大模型 / 代码执行 / 第三方 API应用场景:校对润色 / 模板 / 研究

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add promptify
Metadata
name
promptify
description
Optimize prompts for clarity and effectiveness. Use when user says "improve this prompt", "optimize my prompt", "make this clearer", or provides vague/unstructured prompts. Intelligently routes to sub-agents for codebase research, clarifying questions, or web search as needed.
metadata
{"moltbot":{"emoji":"✨"}}

Prompt Optimizer

Transform prompts into clear, effective ones. Model-agnostic.

Modifiers (parse from ARGUMENTS)

  • +ask → Force clarifying questions
  • +deep → Force codebase exploration
  • +web → Force web search

No modifiers? Auto-detect what's needed.

Auto-Detection Triggers

Trigger Signals
codebase-researcher "this project", "our API", specific files/functions, "integrate", "extend", "refactor"
clarifier Ambiguous ("make it better"), multiple interpretations, missing constraints, vague pronouns
web-researcher "best practices", "latest", external APIs/libraries, framework patterns, year references

Agent Dispatch

When agents needed:

  1. Announce which and why
  2. Run in parallel via Task tool (agents/ directory)
  3. Synthesize findings
  4. Optimize with gathered context

Core Contract (every prompt needs all four)

Element If Missing
Role Add persona with expertise
Task Make action specific
Constraints Infer from context
Output Specify format/structure

Process

  1. If image: Analyze, incorporate context
  2. Detect type: coding/writing/analysis/creative/data
  3. Convert output→process: "Write X" → "Analyze → Plan → Implement → Validate"
  4. Strip fluff: "please", "I want you to", filler, apologies
  5. Apply contract: Verify all 4 elements
  6. Add structure: XML tags for complex prompts

Type Focus

  • Coding: Specs, edge cases, framework
  • Writing: Tone, audience, length
  • Analysis: Criteria, depth
  • Creative: Constraints, novelty
  • Data: I/O format, edge cases

Output

  1. Optimized prompt in code block
  2. echo 'PROMPT' | pbcopy
  3. 2-3 sentence explanation