Skip to content
Back to search

prompt-architect

Agent skill for analyzing and improving prompts using 31 frameworks across 7 intent categories. Works with Claude Code, Gemini CLI, Cursor, Copilot, and 30+ Agent Skills compatible tools.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/ckelsoe--prompt-architect/badge.svg)](/listings/ckelsoe--prompt-architect)

HTML

<a href="/listings/ckelsoe--prompt-architect"><img src="/api/listings/ckelsoe--prompt-architect/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

npx @ckelsoe/prompt-architect

The interactive installer detects your AI agents (Claude Code, Gemini CLI, Cursor, Copilot, Codex, and more) and lets you choose where to install.

> **Tip:** Use `npx` rather than `npm install`. Both work — `npm install` runs the same installer via the postinstall hook and installs to every agent it detects — but only `npx` gives you the interactive menu to choose specific targets.

> Published on the public npm registry. **No authentication required.**

---

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 20 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "prompt-architect-marketplace",
  3. "owner": {
  4. "name": "Charles Kelsoe",
  5. "email": "charles@kelsoe.com"
  6. },
  7. "metadata": {
  8. "description": "Prompt engineering skill using 31 frameworks across 7 intent categories"
  9. },
  10. "plugins": [
  11. {
  12. "name": "prompt-architect",
  13. "source": "./",
  14. "description": "Analyzes and improves prompts using 31 frameworks. Intent-based selection routes to the right framework — from ultra-minimal (APE) to agentic (ReAct).",
  15. "version": "3.5.1",
  16. "keywords": ["prompt-engineering", "prompt-frameworks", "agent-skills"],
  17. "category": "productivity"
  18. }
  19. ]
  20. }
README.md · README · 1,072 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Prompt Architect
  2. Transform vague prompts into expert-level, structured prompts using 31 frameworks across 7 intent categories.
  3. Works with **Claude Code, ChatGPT, Gemini CLI, Cursor, GitHub Copilot, Windsurf, OpenAI Codex**, and [30+ Agent Skills compatible tools](https://agentskills.io).
  4. [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![npm](https://img.shields.io/npm/v/@ckelsoe/prompt-architect)](https://www.npmjs.com/package/@ckelsoe/prompt-architect) [![Agent Skills](https://img.shields.io/badge/Agent%20Skills-compatible-blue)](https://agentskills.io) [![Available on CodeGuilds](https://img.shields.io/badge/Available_on-CodeGuilds-6366f1)](https://codeguilds.dev/packages/prompt-architect)
  5. ## Quick Start
  6. ```bash
  7. npx @ckelsoe/prompt-architect
  8. ```
  9. The interactive installer detects your AI agents (Claude Code, Gemini CLI, Cursor, Copilot, Codex, and more) and lets you choose where to install.
  10. > **Tip:** Use `npx` rather than `npm install`. Both work — `npm install` runs the same installer via the postinstall hook and installs to every agent it detects — but only `npx` gives you the interactive menu to choose specific targets.
  11. > Published on the public npm registry. **No authentication required.**
  12. ---
  13. ## Table of Contents
  14. - [Overview](#overview)
  15. - [Key Features](#key-features)
  16. - [Example Transformation](#example-transformation)
  17. - [Supported Frameworks](#supported-frameworks)
  18. - [Quick Start](#quick-start)
  19. - [Installation](#installation)
  20. - [Verifying Your Installation](#verifying-your-installation)
  21. - [Updating](#updating)
  22. - [Usage](#usage)
  23. - [Framework Selection Guide](#framework-selection-guide)
  24. - [Project Structure](#project-structure)
  25. - [Documentation](#documentation)
  26. - [Contributing](#contributing)
  27. - [Support](#support)
  28. - [License](#license)
  29. ---
  30. ## Overview
  31. Prompt Architect is an [Agent Skills](https://agentskills.io) compatible skill that elevates your prompting capabilities through:
  32. - **Intelligent Analysis** - Evaluates prompts across 5 quality dimensions (clarity, specificity, context, completeness, structure)
  33. - **Framework Recommendation** - Suggests the best framework(s) for your specific use case with clear reasoning
  34. - **Guided Dialogue** - Asks targeted clarifying questions to gather missing information progressively
  35. - **Systematic Application** - Applies selected framework to transform your prompt
  36. - **Iterative Refinement** - Continues improving based on feedback until perfect
  37. **Target Audience:**
  38. - Developers using AI coding agents (Claude Code, Gemini CLI, Cursor, Copilot, etc.)
  39. - Prompt engineers optimizing LLM interactions
  40. - AI practitioners seeking systematic prompt improvement
  41. - Teams wanting consistent, high-quality prompts
  42. ---
  43. ## Key Features
  44. ### 31 Frameworks Across 7 Intent Categories
  45. | Framework | Best For | Complexity |
  46. |-----------|----------|------------|
  47. | **CO-STAR** | Content creation, writing tasks | High |
  48. | **RISEN** | Multi-step processes, procedures | High |
  49. | **CRISPE** | Comprehensive prompts with multiple output variants | High |
  50. | **BROKE** | Business deliverables with OKR-style measurable outcomes | Medium |
  51. | **RISE-IE** | Data analysis, transformations (Input-Expectation) | Medium |
  52. | **RISE-IX** | Content creation with examples (Instructions-Examples) | Medium |
  53. | **TIDD-EC** | High-precision tasks with explicit dos/don'ts | Medium |
  54. | **RACE** | Expert tasks requiring role + context + outcome clarity | Medium |
  55. | **CARE** | Constraint-driven tasks with explicit rules and examples | Medium |
  56. | **CTF** | Simple tasks where situational context drives the prompt | Low |
  57. | **RTF** | Simple, focused tasks where expertise framing matters | Low |
  58. | **APE** | Ultra-minimal one-off prompts | Low |
  59. | **BAB** | Rewriting, refactoring, transforming existing content | Low |
  60. | **Tree of Thought** | Decisions requiring exploration of multiple approaches | Medium |
  61. | **ReAct** | Agentic / tool-use tasks with iterative reasoning | Medium |
  62. | **Skeleton of Thought** | Structured long-form content (outline-first) | Medium |
  63. | **Step-Back** | Principle-grounded reasoning (abstract first, then specific) | Medium |
  64. | **Least-to-Most** | Compositional multi-hop problems (simplest first) | Medium |
  65. | **Plan-and-Solve (PS+)** | Zero-shot numerical/calculation reasoning | Low |
  66. | **Chain of Thought** | Reasoning, problem-solving | Medium |
  67. | **Self-Consistency** | Robust answers: sample many reasoning paths, majority-vote | Medium |
  68. | **Chain of Density** | Summarization at fixed length, maximum information density | Medium |
  69. | **Iterative Compression** | Shortening text toward a target length | Low |
  70. | **Self-Refine** | Iterative output quality improvement (any task) | Medium |
  71. | **CAI Critique-Revise** | Principle-based critique and revision (Anthropic) | Medium |
  72. | **Devil's Advocate** | Strongest opposing argument against a position | Low |
  73. | **Pre-Mortem** | Assume failure, identify specific causes | Low |
  74. | **RCoT** | Verify reasoning by reconstructing the question | Medium |
  75. | **Chain-of-Verification** | Fact-check a draft by verifying each claim independently | Medium |
  76. | **RPEF** | Recover/reconstruct a prompt from an existing output | Low |
  77. | **Reverse Role Prompting** | AI interviews you before executing | Low |
  78. **Composable technique** (layered onto any framework, not one of the 31): **Few-shot / in-context examples** — the discipline of adding well-chosen, well-ordered worked examples to a prompt.
  79. ### Quality Scoring System
  80. Every prompt is evaluated across:
  81. - **Clarity** (1-10): Is the goal clear and unambiguous?
  82. - **Specificity** (1-10): Are requirements detailed enough?
  83. - **Context** (1-10): Is background information provided?
  84. - **Completeness** (1-10): Are all necessary elements present?
  85. - **Structure** (1-10): Is the prompt well-organized?
  86. ### Progressive Disclosure
  87. - Asks 3-5 targeted questions at a time (never overwhelming)
  88. - Adapts questions based on framework selection
  89. - Builds understanding iteratively
  90. - Confirms assumptions before proceeding
  91. ### Continuous Iteration
  92. - Shows before/after comparisons with explanations
  93. - Refines based on your feedback
  94. - Can switch frameworks if initial choice doesn't fit
  95. - Continues until you're satisfied
  96. ---
  97. ## Example Transformation
  98. ### Before
  99. ```
  100. "Write about machine learning"
  101. ```
  102. **Analysis Scores:**
  103. - Clarity: 2/10 (vague goal)
  104. - Specificity: 1/10 (no details)
  105. - Context: 0/10 (no background)
  106. - Completeness: 2/10 (missing most elements)
  107. - Structure: 3/10 (single sentence)
  108. - **Overall: 1.6/10**
  109. ---
  110. ### After (CO-STAR Framework)
  111. ```
  112. CONTEXT:
  113. Creating content for a business blog aimed at C-level executives exploring
  114. how AI/ML could benefit their organizations. Readers understand business
  115. strategy but have limited technical ML knowledge. Part of an emerging
  116. technologies series.
  117. OBJECTIVE:
  118. Create an engaging article helping executives understand practical machine
  119. learning applications relevant to their companies. Focus on demonstrating
  120. tangible business value and real-world implementation without overwhelming
  121. technical details.
  122. STYLE:
  123. Professional blog style combining narrative with bullet points. Include 2-3
  124. real-world case studies. Structure with clear subheadings every 150-200 words.
  125. Balance storytelling with concrete information. Avoid jargon; when necessary,
  126. provide plain-language explanations.
  127. TONE:
  128. Professional yet approachable and conversational. Confident and authoritative
  129. without being condescending. Practical and business-focused rather than
  130. theoretical.
  131. AUDIENCE:
  132. C-suite executives and senior managers at mid-to-large enterprises who:
  133. - Make strategic technology investment decisions
  134. - Understand business metrics and ROI
  135. - Have limited technical ML knowledge
  136. - Value practical examples over theory
  137. RESPONSE FORMAT:
  138. 800-word article structured as:
  139. - Compelling headline (10 words max)
  140. - Brief hook (2-3 sentences)
  141. - 3-4 main sections with descriptive subheadings
  142. - Mix of paragraphs and bullet points
  143. - Clear call-to-action conclusion
  144. ```
  145. **Result Scores:**
  146. - Clarity: 9/10
  147. - Specificity: 9/10
  148. - Context: 10/10
  149. - Completeness: 9/10
  150. - Structure: 9/10
  151. - **Overall: 8.8/10**
  152. ---
  153. ## Supported Frameworks
  154. ### CO-STAR (Context, Objective, Style, Tone, Audience, Response)
  155. **Best for:** Content creation, writing tasks, communications
  156. **Components:**
  157. - **Context** - Background information and situation
  158. - **Objective** - Clear goal and purpose
  159. - **Style** - Writing style and formatting approach
  160. - **Tone** - Voice and emotional quality
  161. - **Audience** - Target reader characteristics
  162. - **Response** - Expected format and structure
  163. **Example Use Cases:** Blog posts, emails, presentations, marketing copy, documentation
  164. ---
  165. ### RISEN (Role, Instructions, Steps, End goal, Narrowing)
  166. **Best for:** Multi-step processes, systematic procedures
  167. **Components:**
  168. - **Role** - Expertise and perspective needed
  169. - **Instructions** - High-level guidance
  170. - **Steps** - Detailed methodology
  171. - **End goal** - Success criteria
  172. - **Narrowing** - Constraints and boundaries
  173. **Example Use Cases:** Code reviews, workflows, systematic analysis, project planning
  174. ---
  175. ### RISE (Dual Variants)
  176. #### RISE-IE (Input-Expectation)
  177. **Best for:** Data analysis, transformations, processing tasks
  178. **Components:**
  179. - **Role** - Expertise needed
  180. - **Input** - Data format and characteristics
  181. - **Steps** - Processing methodology
  182. - **Expectation** - Output requirements
  183. **Example Use Cases:** CSV analysis, data processing, file transformations, report generation
  184. #### RISE-IX (Instructions-Examples)
  185. **Best for:** Content creation with reference examples
  186. **Components:**
  187. - **Role** - Expertise needed
  188. - **Instructions** - Task guidance
  189. - **Steps** - Workflow process
  190. - **Examples** - Reference samples
  191. **Example Use Cases:** Creative writing, template-based content, style matching
  192. ---
  193. ### TIDD-EC (Task, Instructions, Do, Don't, Examples, Context)
  194. **Best for:** High-precision tasks requiring explicit boundaries
  195. **Components:**
  196. - **Task type** - Nature of the work
  197. - **Instructions** - What to accomplish
  198. - **Do** - Explicit positive guidance
  199. - **Don't** - Explicit negative guidance (what to avoid)
  200. - **Examples** - Reference samples
  201. - **Context** - Background information
  202. **Example Use Cases:** Code generation with standards, compliance tasks, quality-critical work
  203. ---
  204. ### CTF (Context, Task, Format)
  205. **Best for:** Simple tasks where situational background matters more than expertise framing
  206. **Components:**
  207. - **Context** - Situation and background
  208. - **Task** - What needs to be done
  209. - **Format** - Output structure
  210. **Example Use Cases:** Handoff documents, mid-project updates, situation-driven requests
  211. ---
  212. ### RTF (Role, Task, Format)
  213. **Best for:** Simple, well-defined tasks where expertise framing drives output quality
  214. **Components:**
  215. - **Role** - Expertise required
  216. - **Task** - What needs to be done
  217. - **Format** - Output structure
  218. **Example Use Cases:** Quick conversions, simple formatting, straightforward requests
  219. ---
  220. ### APE (Action, Purpose, Expectation)
  221. **Best for:** Ultra-minimal prompts — the simplest structured framework
  222. **Components:**
  223. - **Action** - What to do (one clear verb-driven instruction)
  224. - **Purpose** - Why it's needed (one sentence)
  225. - **Expectation** - What a good result looks like
  226. **Example Use Cases:** Quick summaries, single-function code, one-off requests, rapid iteration
  227. ---
  228. ### BAB (Before, After, Bridge)
  229. **Best for:** Transforming, rewriting, or refactoring existing content
  230. **Components:**
  231. - **Before** - Current state and what's wrong
  232. - **After** - Desired end state
  233. - **Bridge** - Transformation rules and constraints
  234. **Example Use Cases:** Code refactoring, copy rewrites, tone changes, document restructuring, version migrations
  235. ---
  236. ### RACE (Role, Action, Context, Expectation)
  237. **Best for:** Medium-complexity tasks needing expertise + background + explicit success criteria
  238. **Components:**
  239. - **Role** - Expertise or persona
  240. - **Action** - What needs to be done
  241. - **Context** - Situational background and constraints
  242. - **Expectation** - What a good result looks like
  243. **Example Use Cases:** Technical reviews, expert analysis, contextual recommendations, documentation with standards
  244. ---
  245. ### CRISPE (Capacity+Role, Insight, Instructions, Personality, Experiment)
  246. **Best for:** Comprehensive prompts where you want multiple output variants to compare
  247. **Components:**
  248. - **Capacity & Role** - Expertise level and professional persona
  249. - **Insight** - Background context and situational knowledge
  250. - **Instructions** - The specific task
  251. - **Personality** - Tone, voice, communication style
  252. - **Experiment** - Request N variants to compare (the defining differentiator)
  253. **Example Use Cases:** Marketing campaigns (A/B variants), content with tone options, strategic analysis needing multiple angles
  254. ---
  255. ### BROKE (Background, Role, Objective, Key Results, Evolve)
  256. **Best for:** Business deliverables with measurable outcomes and built-in self-improvement
  257. **Components:**
  258. - **Background** - Situation and context
  259. - **Role** - Professional persona
  260. - **Objective** - The task
  261. - **Key Results** - Measurable business outcomes (OKR-style)
  262. - **Evolve** - AI self-critiques and suggests 3 improvements
  263. **Example Use Cases:** Sales process improvements, content strategy with KPIs, product decisions tied to metrics
  264. ---
  265. ### CARE (Context, Ask, Rules, Examples)
  266. **Best for:** Tasks with explicit constraints, compliance requirements, or quality standards
  267. **Components:**
  268. - **Context** - Situation and background
  269. - **Ask** - The specific request
  270. - **Rules** - Explicit constraints, dos/don'ts, standards
  271. - **Examples** - Reference samples showing the quality bar
  272. **Example Use Cases:** Healthcare/legal content, UI error messages, interview questions with bias constraints, brand-compliant copy
  273. ---
  274. ### Skeleton of Thought (SoT)
  275. **Best for:** Structured long-form content — generate outline first, then expand
  276. **Approach:**
  277. - Phase 1: Generate a concise skeleton/outline (key points only)
  278. - Phase 2: Expand each skeleton point independently (parallelizable)
  279. - SoT-R variant: Route non-suitable questions to standard generation
  280. **Example Use Cases:** Technical documentation, structured reports, tutorials, any multi-section content
  281. ---
  282. ### Step-Back Prompting
  283. **Best for:** Principle-grounded reasoning — abstract to the underlying concept first
  284. **Approach:**
  285. - Generate a higher-level "step-back" question about underlying principles
  286. - Answer the abstract question to retrieve principles
  287. - Use those principles as context to answer the original specific question
  288. **Example Use Cases:** STEM problems, architecture decisions, debugging, any task where first-principles reasoning matters
  289. ---
  290. ### Least-to-Most (LtM)
  291. **Best for:** Compositional multi-hop problems with ordered dependencies
  292. **Approach:**
  293. - Decompose the complex problem into ordered subproblems (simplest first)
  294. - Solve each subproblem sequentially, feeding each answer into the next
  295. - Use accumulated answers to solve the original problem
  296. **Example Use Cases:** Multi-domain questions (legal + technical), complex calculations, architecture problems with prerequisites
  297. ---
  298. ### Plan-and-Solve (PS+)
  299. **Best for:** Zero-shot numerical and calculation reasoning
  300. **Approach:**
  301. - "Let's first understand the problem, extract relevant variables and their corresponding numerals, and devise a complete plan. Then, let's carry out the plan, calculate intermediate values, pay attention to computation, and solve the problem step by step."
  302. **Example Use Cases:** Financial calculations (MRR, CAC, payback), math word problems, resource estimation, any zero-shot reasoning task
  303. ---
  304. ### Self-Refine
  305. **Best for:** Iterative quality improvement of any output
  306. **Approach:** Generate → produce specific actionable feedback → refine → repeat until stopping criterion
  307. **Research:** Madaan et al. NeurIPS 2023 — +5-40% improvement across 7 task types
  308. **Example Use Cases:** Code review and rewriting, writing improvement, plan refinement, pre-submission QA
  309. ---
  310. ### CAI Critique-Revise
  311. **Best for:** Aligning output to an explicit stated principle or standard
  312. **Approach:** Initial output → critique against a specific principle → revision addressing every critique point
  313. **Research:** Anthropic Constitutional AI (arXiv 2212.08073, 2022) — principle-driven alignment
  314. **Example Use Cases:** Plain language compliance, brand voice enforcement, epistemic quality (claims vs. assertions), legal/regulatory language standards
  315. ---
  316. ### Devil's Advocate
  317. **Best for:** Generating the strongest possible opposing argument against a position
  318. **Approach:** Explicitly instructs the AI to attack a position as forcefully as possible — not balanced, not a straw man, but maximum-strength opposition
  319. **Research:** ACM IUI 2024 peer-reviewed study
  320. **Example Use Cases:** Decision stress-testing, architecture reviews, countering groupthink, stakeholder preparation, debiasing
  321. ---
  322. ### Pre-Mortem
  323. **Best for:** Identifying specific failure causes before they happen
  324. **Approach:** Assume the project has already failed → describe the failure → work backwards to specific causes with warning signs
  325. **Research:** Mitchell, Russo & Pennington (1989), popularized by Gary Klein (HBR 2007) — prospective hindsight improves identification of reasons for future outcomes by ~30%; LLM application is practitioner-level, with no dedicated AI paper
  326. **Example Use Cases:** Project kickoffs, product launches, technical migrations, high-stakes strategic decisions
  327. ---
  328. ### RCoT (Reverse Chain-of-Thought)
  329. **Best for:** Verifying that reasoning addressed all conditions in a multi-constraint question
  330. **Approach:** Generate answer → reconstruct the question from the answer → cross-check conditions → correct overlooked items
  331. **Research:** Academic backward reasoning literature (ACL 2025, NAACL 2025)
  332. **Example Use Cases:** Multi-condition logic problems, complex requirements analysis, high-stakes reasoning verification
  333. ---
  334. ### RPEF (Reverse Prompt Engineering)
  335. **Best for:** Recovering a reusable prompt template from an existing output
  336. **Approach:** Provide an output (and optionally the input) → AI analyzes it for tone, structure, constraints, persona → generates a reusable template with [PLACEHOLDER] variables
  337. **Research:** Li & Klabjan, EMNLP 2025 (arXiv 2411.06729)
  338. **Example Use Cases:** Recovering lost prompts, codifying successful one-time outputs, building style templates, understanding system prompt behavior
  339. ---
  340. ### Reverse Role Prompting (AI-Led Interview)
  341. **Best for:** Complex tasks where you know the goal but struggle to specify all requirements
  342. **Approach:** Provide a minimal intent statement → AI asks targeted clarifying questions → executes once context is complete
  343. **Research:** FATA framework (arXiv 2508.08308, 2025) — ~40% improvement over standard prompting
  344. **Example Use Cases:** Complex strategy tasks, non-expert users, requirements gathering, generating complete prompts from an interview
  345. ---
  346. ### Tree of Thought
  347. **Best for:** Decisions where multiple approaches need systematic comparison
  348. **Approach:**
  349. - Defines the problem and constraints
  350. - Explores 2-5 distinct solution branches in parallel
  351. - Evaluates each branch against defined criteria
  352. - Synthesizes into a reasoned recommendation
  353. **Example Use Cases:** Architecture decisions, debugging with multiple hypotheses, technology selection, strategic trade-offs
  354. ---
  355. ### ReAct (Reasoning + Acting)
  356. **Best for:** Agentic tasks that interleave reasoning with tool use
  357. **Approach:**
  358. - Defines goal, available tools, and constraints
  359. - Alternates Thought → Action → Observation cycles
  360. - Each observation informs the next thought
  361. - Continues until goal is reached
  362. **Example Use Cases:** Agentic workflows, multi-step research, debugging with tools, data investigation
  363. ---
  364. ### Chain of Thought
  365. **Best for:** Complex reasoning and problem-solving
  366. **Approach:**
  367. - Breaks down reasoning into explicit steps
  368. - Shows work and intermediate conclusions
  369. - Verifies logic at each stage
  370. - Builds to final answer
  371. **Example Use Cases:** Math problems, debugging, decision analysis, logical reasoning
  372. ---
  373. ### Chain of Density
  374. **Best for:** Iterative refinement and compression
  375. **Approach:**
  376. - Starts with baseline version
  377. - Progressively refines through iterations
  378. - Increases information density
  379. - Optimizes for specific goals
  380. **Example Use Cases:** Summarization, content compression, explanation optimization
  381. ---
  382. ## Installation
  383. Choose the method that matches your AI tool:
  384. ### Claude Code
  385. ```
  386. /install-skill https://github.com/ckelsoe/prompt-architect/tree/main/skills/prompt-architect
  387. ```
  388. **Plugin system (enables updates):**
  389. ```
  390. /plugin marketplace add ckelsoe/prompt-architect
  391. /plugin install prompt-architect@prompt-architect-marketplace
  392. ```
  393. ### Codex CLI
  394. ```
  395. $skill-installer install https://github.com/ckelsoe/prompt-architect/tree/main/skills/prompt-architect
  396. ```
  397. ### Other Agents (Gemini, Cursor, Copilot, Cline, Zoo Code, etc.)
  398. Copy `skills/prompt-architect/` from this repo to `~/.agents/skills/prompt-architect/`
  399. ### npm (alternative)
  400. ```bash
  401. npx @ckelsoe/prompt-architect
  402. ```
  403. Installs to both `.claude/skills/` and `.agents/skills/` in one step.
  404. ### ChatGPT
  405. ChatGPT natively supports [Agent Skills](https://agentskills.io). Available on Business, Enterprise, Edu, Teachers, and Healthcare plans.
  406. 1. Download [`prompt-architect.zip`](https://github.com/ckelsoe/prompt-architect/releases/latest/download/prompt-architect.zip) (direct download, always latest version)
  407. 2. In ChatGPT, click your profile icon → **Skills**
  408. 3. Click **New skill** → **Upload from your computer**
  409. 4. Upload the `.zip` file — do not unzip it first
  410. ChatGPT scans uploaded skills before making them available; this one is plain
  411. Markdown with no scripts, so the scan is typically immediate.
  412. > Use the `.zip`, not the `.skill` file also attached to the release. They are
  413. > the same archive, but ChatGPT's uploader expects a `.zip` containing a single
  414. > top-level folder.
  415. ### Windsurf
  416. Windsurf does not yet support Agent Skills natively:
  417. ```bash
  418. cat adapters/for-windsurf.md >> .windsurfrules
  419. ```
  420. ### Any LLM API or Chat Interface
  421. Paste the contents of [`adapters/system-prompt.md`](https://raw.githubusercontent.com/ckelsoe/prompt-architect/main/adapters/system-prompt.md) into your system prompt field. Works with OpenAI API, Anthropic API, Google Gemini API, Custom GPTs, or any tool with a system prompt.
  422. See [`adapters/README.md`](adapters/README.md) for detailed instructions for each platform.
  423. ---
  424. ## Verifying Your Installation
  425. After installing, test with:
  426. ```
  427. "Help me improve this prompt: write a technical blog post"
  428. ```
  429. The skill will automatically:
  430. 1. Analyze your prompt and score it
  431. 2. Recommend the best framework (likely CO-STAR)
  432. 3. Ask clarifying questions
  433. 4. Generate an improved, structured prompt
  434. 5. Explain the changes made
  435. ---
  436. ## Updating
  437. ```bash
  438. # npm
  439. npm update -g @ckelsoe/prompt-architect
  440. # Check version
  441. npm list -g @ckelsoe/prompt-architect
  442. # Claude Code plugin
  443. /plugin update prompt-architect@prompt-architect-marketplace
  444. ```
  445. > **Migrating from v2.x?** The package was renamed from `@ckelsoe/claude-skill-prompt-architect`. See [MIGRATION.md](MIGRATION.md) for details.
  446. ---
  447. ## Usage
  448. ### Basic Usage
  449. Simply ask Claude Code to improve any prompt:
  450. ```
  451. "Help me improve this prompt: [your prompt here]"
  452. ```
  453. ### Example Interaction
  454. ```
  455. User: "Help me improve this prompt: create a function to process user data"
  456. Skill: I'll analyze this prompt and help improve it.
  457. **Analysis**:
  458. - Clarity: 4/10 (basic goal stated)
  459. - Specificity: 2/10 (what kind of data? what processing?)
  460. - Context: 1/10 (no background)
  461. - Completeness: 2/10 (missing many details)
  462. - Structure: 3/10 (single sentence)
  463. - **Overall: 2.4/10**
  464. **Recommendation**: RISEN framework (multi-step process requiring methodology)
  465. **Questions**:
  466. 1. What type of user data? (names, emails, profiles, etc.)
  467. 2. What processing is needed? (validation, transformation, storage?)
  468. 3. What's the tech stack? (Python, JavaScript, database type?)
  469. 4. Any constraints? (performance, security requirements?)
  470. 5. Expected error handling approach?
  471. User: "It's a Python function processing CSV user registrations. Need to validate
  472. emails, normalize names, and save to PostgreSQL. Should handle invalid data gracefully."
  473. Skill: Perfect! Let me apply the RISEN framework...
  474. [Improved prompt generated with clear role, instructions, steps, end goal, and constraints]
  475. ```
  476. ### Advanced Usage
  477. **Request Specific Framework:**
  478. ```
  479. "Use the CO-STAR framework to improve: write marketing copy"
  480. ```
  481. **Iterate on Existing Prompts:**
  482. ```
  483. "Review this prompt and suggest improvements: [existing structured prompt]"
  484. ```
  485. **Switch Frameworks:**
  486. ```
  487. "Try using TIDD-EC instead for more explicit guidance"
  488. ```
  489. ---
  490. ## Framework Selection Guide
  491. ### Decision Tree
  492. ```
  493. Are you transforming existing content (rewrite, refactor, convert)?
  494. ├─ YES → BAB (Before, After, Bridge)
  495. └─ NO ↓
  496. Is it an agentic/tool-use task?
  497. ├─ YES → ReAct (Thought-Action-Observation cycles)
  498. └─ NO ↓
  499. Is it a decision between multiple approaches?
  500. ├─ YES → Tree of Thought (branching exploration)
  501. └─ NO ↓
  502. Is it content/writing focused with audience and tone?
  503. ├─ YES → CO-STAR
  504. └─ NO ↓
  505. Is it a multi-step process with methodology?
  506. ├─ YES → RISEN
  507. └─ NO ↓
  508. Is it a data transformation (input → output)?
  509. ├─ YES → RISE-IE
  510. └─ NO ↓
  511. Does it need reference examples?
  512. ├─ YES → RISE-IX
  513. └─ NO ↓
  514. Does it need explicit dos/don'ts?
  515. ├─ YES → TIDD-EC
  516. └─ NO ↓
  517. Is it complex reasoning (one clear path)?
  518. ├─ YES → Chain of Thought
  519. └─ NO ↓
  520. Does it need iterative refinement?
  521. ├─ YES → Chain of Density
  522. └─ NO ↓
  523. Is it a numerical/calculation problem (zero-shot)?
  524. ├─ YES → Plan-and-Solve PS+
  525. └─ NO ↓
  526. Is it compositional / multi-hop (answer A needed before B)?
  527. ├─ YES → Least-to-Most
  528. └─ NO ↓
  529. Does it need first-principles reasoning?
  530. ├─ YES → Step-Back Prompting
  531. └─ NO ↓
  532. Is it structured long-form content (multiple semi-independent sections)?
  533. ├─ YES → Skeleton of Thought
  534. └─ NO ↓
  535. Do you need to recover a prompt from an existing output?
  536. ├─ YES → RPEF (Reverse Prompt Engineering)
  537. └─ NO ↓
  538. Do you need to clarify requirements before starting?
  539. ├─ YES → Reverse Role Prompting (AI interviews you first)
  540. └─ NO ↓
  541. Do you need to critique, stress-test, or verify something?
  542. ├─ General quality improvement → Self-Refine
  543. ├─ Align to explicit principle → CAI Critique-Revise
  544. ├─ Find opposing argument → Devil's Advocate
  545. ├─ Find failure modes → Pre-Mortem
  546. └─ Verify conditions weren't missed → RCoT
  547. Is it a simple task? Choose by primary driver:
  548. ├─ Expert role matters most → RTF
  549. ├─ Situational context matters most → CTF
  550. ├─ Need role + context + outcome bar → RACE
  551. ├─ Business deliverable with KPIs → BROKE
  552. ├─ Want multiple variants to compare → CRISPE
  553. ├─ Have explicit rules/constraints → CARE
  554. └─ Ultra-minimal, one-off → APE
  555. ```
  556. ### Quick Reference
  557. | Your Task | Recommended Framework | Why |
  558. |-----------|----------------------|-----|
  559. | Write blog post | CO-STAR | Audience and tone critical |
  560. | Code review process | RISEN | Multi-step with constraints |
  561. | Analyze CSV data | RISE-IE | Clear input→output transformation |
  562. | Generate with examples | RISE-IX | Need reference samples |
  563. | Generate secure code | TIDD-EC | Explicit dos/don'ts needed |
  564. | Debug logic error | Chain of Thought | Requires reasoning steps |
  565. | Compress explanation | Chain of Density | Iterative refinement |
  566. | Simple conversion | RTF | Straightforward, expertise-driven |
  567. | Mid-project update | CTF | Background is the key driver |
  568. | Summarize a meeting | APE | Ultra-minimal, one-off |
  569. | Refactor existing code | BAB | Clear before/after transformation |
  570. | Rewrite copy for new audience | BAB | Current content → desired state |
  571. | Architecture decision | Tree of Thought | Multiple approaches to compare |
  572. | Choose database tech | Tree of Thought | Trade-offs need systematic analysis |
  573. | Agentic research task | ReAct | Tool use with iterative reasoning |
  574. | Expert review with context | RACE | Role + background + outcome bar |
  575. | Financial calculation | Plan-and-Solve (PS+) | Zero-shot + variable extraction |
  576. | Multi-hop technical question | Least-to-Most | Dependencies must be solved in order |
  577. | Architecture principle question | Step-Back | Abstract to first principles first |
  578. | Write structured docs/report | Skeleton of Thought | Outline first, expand second |
  579. | Marketing copy A/B options | CRISPE | Experiment component generates variants |
  580. | Business strategy with KPIs | BROKE | Key Results + Evolve self-critique |
  581. | Healthcare/compliance content | CARE | Explicit rules and quality standards |
  582. | Improve any output quality | Self-Refine | Multi-dimensional feedback + refine loop |
  583. | Enforce plain language standard | CAI Critique-Revise | Principle-based critique |
  584. | Stress-test a strategy | Devil's Advocate | Strongest opposing argument |
  585. | Project risk analysis | Pre-Mortem | Assumed failure → backward causes |
  586. | Multi-condition logic verification | RCoT | Backward reconstruction cross-check |
  587. | Lost/need to recover a prompt | RPEF | Output → reconstructed template |
  588. | Requirements unclear upfront | Reverse Role Prompting | AI interviews you first |
  589. ---
  590. ## Project Structure
  591. ```
  592. prompt-architect/
  593. │
  594. ├── README.md # This file
  595. ├── LICENSE # MIT License
  596. │
  597. └── prompt-architect/ # The skill
  598. ├── SKILL.md # Core skill instructions
  599. │
  600. ├── references/ # Framework Documentation
  601. │ └── frameworks/ # Loaded on-demand
  602. │ ├── co-star.md # CO-STAR reference (600+ lines)
  603. │ ├── risen.md # RISEN reference (600+ lines)
  604. │ ├── rise.md # RISE (IE/IX) reference (700+ lines)
  605. │ ├── tidd-ec.md # TIDD-EC reference (600+ lines)
  606. │ ├── ctf.md # CTF reference
  607. │ ├── rtf.md # RTF reference (500+ lines)
  608. │ ├── ape.md # APE reference
  609. │ ├── bab.md # BAB reference
  610. │ ├── race.md # RACE reference
  611. │ ├── crispe.md # CRISPE reference
  612. │ ├── broke.md # BROKE reference
  613. │ ├── care.md # CARE reference
  614. │ ├── tree-of-thought.md # Tree of Thought reference
  615. │ ├── react.md # ReAct reference
  616. │ ├── skeleton-of-thought.md # Skeleton of Thought reference (ICLR 2024)
  617. │ ├── step-back.md # Step-Back Prompting reference (Google DeepMind)
  618. │ ├── least-to-most.md # Least-to-Most reference (Google Brain)
  619. │ ├── plan-and-solve.md # Plan-and-Solve PS+ reference (ACL 2023)
  620. │ ├── chain-of-thought.md # CoT reference (500+ lines)
  621. │ ├── chain-of-density.md # CoD reference (500+ lines)
  622. │ ├── self-refine.md # Self-Refine (NeurIPS 2023)
  623. │ ├── cai-critique-revise.md # CAI Critique-Revise (Anthropic 2022)
  624. │ ├── devils-advocate.md # Devil's Advocate (ACM IUI 2024)
  625. │ ├── pre-mortem.md # Pre-Mortem (Gary Klein)
  626. │ ├── rcot.md # Reverse Chain-of-Thought
  627. │ ├── rpef.md # Reverse Prompt Engineering (EMNLP 2025)
  628. │ └── reverse-role.md # AI-Led Interview / FATA (arXiv 2025)
  629. │
  630. └── assets/
  631. └── templates/ # Framework Templates
  632. ├── co-star_template.txt
  633. ├── risen_template.txt
  634. ├── rise-ie_template.txt
  635. ├── rise-ix_template.txt
  636. ├── tidd-ec_template.txt
  637. ├── ctf_template.txt
  638. ├── rtf_template.txt
  639. ├── ape_template.txt
  640. ├── bab_template.txt
  641. ├── race_template.txt
  642. ├── crispe_template.txt
  643. ├── broke_template.txt
  644. ├── care_template.txt
  645. ├── tree-of-thought_template.txt
  646. ├── react_template.txt
  647. ├── skeleton-of-thought_template.txt
  648. ├── step-back_template.txt
  649. ├── least-to-most_template.txt
  650. ├── plan-and-solve_template.txt
  651. ├── chain-of-thought_template.txt
  652. ├── chain-of-density_template.txt
  653. ├── self-refine_template.txt
  654. ├── cai-critique-revise_template.txt
  655. ├── devils-advocate_template.txt
  656. ├── pre-mortem_template.txt
  657. ├── rcot_template.txt
  658. ├── rpef_template.txt
  659. ├── reverse-role_template.txt
  660. └── hybrid_template.txt
  661. ```
  662. **Core Components:**
  663. - **SKILL.md** - Main skill logic and instructions (intent-based selection system)
  664. - **Framework Reference Docs** - 30 files covering all 31 frameworks, with examples (RISE documents two variants)
  665. - **Templates** - 30 ready-to-use structures: one per framework, plus `hybrid_template.txt` for combinations
  666. ---
  667. ## Documentation
  668. ### Framework References
  669. Detailed documentation for each framework in `prompt-architect/references/frameworks/`:
  670. - **co-star.md** - Context, Objective, Style, Tone, Audience, Response
  671. - **risen.md** - Role, Instructions, Steps, End goal, Narrowing
  672. - **rise.md** - Dual variants (IE & IX)
  673. - **tidd-ec.md** - Task, Instructions, Do, Don't, Examples, Context
  674. - **ctf.md** - Context, Task, Format
  675. - **rtf.md** - Role, Task, Format
  676. - **ape.md** - Action, Purpose, Expectation
  677. - **bab.md** - Before, After, Bridge
  678. - **race.md** - Role, Action, Context, Expectation
  679. - **crispe.md** - Capacity+Role, Insight, Instructions, Personality, Experiment
  680. - **broke.md** - Background, Role, Objective, Key Results, Evolve
  681. - **care.md** - Context, Ask, Rules, Examples
  682. - **tree-of-thought.md** - Branching multi-path exploration
  683. - **react.md** - Reasoning + Acting (agentic tool use)
  684. - **skeleton-of-thought.md** - Skeleton-first parallel expansion (ICLR 2024)
  685. - **step-back.md** - Abstract to principles first (Google DeepMind, ICLR 2024)
  686. - **least-to-most.md** - Sequential subproblem solving (Google Brain, ICLR 2023)
  687. - **plan-and-solve.md** - Zero-shot PS+ trigger (ACL 2023)
  688. - **chain-of-thought.md** - Step-by-step reasoning
  689. - **chain-of-density.md** - Iterative refinement
  690. - **self-refine.md** - Generate → Feedback → Refine (NeurIPS 2023)
  691. - **cai-critique-revise.md** - Principle-based critique + revision (Anthropic 2022)
  692. - **devils-advocate.md** - Strongest opposing argument (ACM IUI 2024)
  693. - **pre-mortem.md** - Assumed failure → backward cause analysis (Gary Klein)
  694. - **rcot.md** - Reverse Chain-of-Thought verification
  695. - **rpef.md** - Reverse Prompt Engineering (EMNLP 2025)
  696. - **reverse-role.md** - AI-Led Interview / FATA (arXiv 2025)
  697. ---
  698. ## Contributing
  699. Contributions are welcome! Here's how you can help:
  700. ### Ways to Contribute
  701. 1. **Add New Frameworks** - Implement additional prompting frameworks
  702. 2. **Improve Existing Frameworks** - Enhance documentation and examples
  703. 3. **Bug Fixes** - Report and fix issues
  704. 4. **Documentation** - Improve guides and examples
  705. 5. **Testing** - Test with different use cases and report findings
  706. ### Adding New Frameworks
  707. 1. Research and validate the framework has backing
  708. 2. Create comprehensive framework reference (see existing in `references/frameworks/`)
  709. 3. Add template file in `assets/templates/`
  710. 4. Update SKILL.md framework selection logic
  711. 5. Test with multiple use cases
  712. 6. Submit pull request with examples and rationale
  713. ### Contribution Guidelines
  714. **File Structure:**
  715. ```
  716. New framework contributions should include:
  717. - references/frameworks/your-framework.md (600+ lines with examples)
  718. - assets/templates/your-framework_template.txt
  719. - Updates to SKILL.md framework selection
  720. - Examples in documentation
  721. ```
  722. **Documentation Standards:**
  723. - Clear component definitions
  724. - Multiple use case examples
  725. - Selection criteria explanation
  726. - Before/after comparisons
  727. - Research citations
  728. **Testing Requirements:**
  729. - Test with 3-5 diverse prompts
  730. - Verify scoring improves
  731. - Check framework selection logic
  732. - Validate template structure
  733. ### Pull Request Process
  734. 1. Fork the repository
  735. 2. Create a feature branch (`git checkout -b feature/new-framework`)
  736. 3. Make your changes following the guidelines
  737. 4. Test thoroughly with multiple examples
  738. 5. Update relevant documentation
  739. 6. Commit with clear messages
  740. 7. Push to your fork
  741. 8. Submit a Pull Request with:
  742. - Clear description of changes
  743. - Examples of improvements
  744. - Test results
  745. - Documentation updates
  746. ---
  747. ## License
  748. This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
  749. ### License Summary
  750. **Permissions:**
  751. - Commercial use
  752. - Modification
  753. - Distribution
  754. - Private use
  755. **Conditions:**
  756. - License and copyright notice must be included
  757. **Limitations:**
  758. - No liability
  759. - No warranty
  760. ---
  761. ## Support
  762. ### Getting Help
  763. - **Issues** - Report bugs or request features via [GitHub Issues](https://github.com/ckelsoe/prompt-architect/issues)
  764. - **Documentation** - Check framework references in `references/frameworks/`
  765. ### FAQ
  766. **Q: Why isn't the skill activating?**
  767. A: Ensure the skill folder is in the correct location (`~/.claude/skills/prompt-architect/`) and restart Claude Code.
  768. **Q: Can I use this without Claude Code?**
  769. A: Yes! Include SKILL.md content in your Claude API system prompt.
  770. **Q: Which framework should I use?**
  771. A: The skill will recommend one, but check the Framework Selection Guide above.
  772. **Q: Can I add my own framework?**
  773. A: Yes! See the Contributing section for guidelines.
  774. **Q: Does this work with other LLMs?**
  775. A: The frameworks are universal, but the skill is optimized for Claude.
  776. ---
  777. **Transforming your prompts with research-backed frameworks**

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.