create-agentic-starter 0.1.3 → 0.1.5

package/README.md

@@ -1,11 +1,53 @@
 # create-agentic-starter
 
-Scaffolds a minimal AI-powered project delivery workflow into the current folder.
+`create-agentic-starter` scaffolds a lightweight `.agentic/` workspace inside an existing project so AI tools can follow a repeatable kick-off workflow.
+
+The goal is to avoid rebuilding the same project-start process every time. Instead of manually re-creating prompts and folders for every new client engagement, this package sets up a standard structure for:
+
+- requirement gathering
+- architecture thinking
+- BRD generation
+- FRD generation
+- estimation
+- proposal drafting
+
+It is designed for prompt-driven AI tools. The generated scaffold gives each new session a clear entrypoint with `@.agentic/init.md`, a set of reusable prompts in `.agentic/commands/`, and a workspace for project assets, memory, and generated documents.
 
 ## Usage
 
+```bash
+npx create-agentic-starter
+```
+
 ```bash
 bunx create-agentic-starter
 ```
 
-The command is non-interactive. It resets the generated scaffold paths, creates the `.agentic/` structure, and prints the next step for the team.
+Run the command inside an existing project folder.
+
+The scaffold is non-interactive. It will:
+
+- recreate `.agentic/`
+- preserve existing `AGENTS.md` and `.gitignore`, appending a marked `create-agentic-starter` section when needed
+- never modify `README.md`
+- print the next step for the team
+
+## Generated Structure
+
+The command sets up:
+
+- `.agentic/init.md`
+- `.agentic/context.md`
+- `.agentic/commands/`
+- `.agentic/workspace/project/`
+- `.agentic/workspace/memory/`
+- `.agentic/workspace/documents/`
+- `AGENTS.md`
+
+## First Run Flow
+
+After scaffolding:
+
+1. Open your AI tool in the project.
+2. Start a new session with `@.agentic/init.md`.
+3. Follow the command sequence defined in `AGENTS.md`.

package/dist/index.js

@@ -1,47 +1,9 @@
 #!/usr/bin/env node
-import{mkdir as f,readFile as h,rm as k,writeFile as c}from"fs/promises";import i from"path";var y="README.md",j=".gitignore",p="<!-- from create-agentic-starter:start -->",l="<!-- from create-agentic-starter:end -->",u="# from create-agentic-starter:start",g="# from create-agentic-starter:end",v=[".agentic","AGENTS.md"],x=()=>({"AGENTS.md":`# Agentic Starter Guide
-
-This scaffold creates a reusable AI working area under \`.agentic/\` for project kick-off and pre-delivery documentation.
-
-## What This Is For
-
-Use this scaffold when you want an AI tool to:
-
-- understand a project from files, notes, screenshots, and links
-- maintain structured project memory
-- generate markdown deliverables such as BRD, FRD, estimate, and proposal
-
-## Folder Purpose
-
-- \`.agentic/init.md\`: first prompt for every new AI session
-- \`.agentic/context.md\`: shared rules and folder map
-- \`.agentic/commands/\`: task-specific prompt files you reference in your AI tool
-- \`.agentic/workspace/project/\`: raw project inputs such as screenshots, notes, requirement docs, and reference links
-- \`.agentic/workspace/memory/\`: evolving AI-maintained project understanding
-- \`.agentic/workspace/documents/\`: polished markdown deliverables
-
-## How To Use It
-
-1. Open your AI tool in this project.
-2. Start every new session with \`@.agentic/init.md\`.
-3. Follow the command sequence below.
-
-## V1 Command Sequence
-
-1. \`@.agentic/init.md\`
-2. \`@.agentic/commands/project-requirements.md\`
-3. \`@.agentic/commands/architecture.md\`
-4. \`@.agentic/commands/create-brd.md\`
-5. \`@.agentic/commands/create-frd.md\`
-6. \`@.agentic/commands/create-estimate.md\`
-7. \`@.agentic/commands/create-proposal.md\`
-
-If your tool does not support \`@file\` references, paste the contents of the prompt file into a new chat manually.
-`,".agentic/init.md":`# Init
+import{mkdir as f,readFile as w,rm as k,writeFile as c}from"fs/promises";import i from"path";var y="AGENTS.md",j=".gitignore",p="<!-- from create-agentic-starter:agents:start -->",u="<!-- from create-agentic-starter:agents:end -->",l="# from create-agentic-starter:start",g="# from create-agentic-starter:end",x=[".agentic"],v=()=>({".agentic/init.md":`# Init
 
 Read \`@.agentic/context.md\` first.
 
-You are starting a fresh AI session inside this project. Your job is to build lightweight orientation without over-reading the repository.
+You are starting a fresh AI session inside this project. Your job is to quickly understand what this project is, what context already exists, and what the user still needs to provide.
 
 ## Read Scope
 
@@ -53,18 +15,39 @@ You are starting a fresh AI session inside this project. Your job is to build li
    - inspect \`.agentic/workspace/documents/\`
 4. Use existing workspace context if it is already present.
 
+## Response Style
+
+1. Be concise and synthesis-first.
+2. Do not dump a full file inventory unless the user explicitly asks.
+3. Do not mirror the prompt structure back to the user.
+4. Respond with either:
+   - a short synthesized summary of what you understand, or
+   - a few focused questions if important context is still missing
+
 ## What To Do
 
-1. Summarize the root-level project structure.
-2. Summarize what already exists in \`.agentic/workspace/project/\`, \`.agentic/workspace/memory/\`, and \`.agentic/workspace/documents/\`.
-3. Build a lightweight understanding from:
-   - root-level structure
+1. Infer what kind of project this is from the root-level signals.
+2. Check whether \`.agentic/workspace/project/\`, \`.agentic/workspace/memory/\`, and \`.agentic/workspace/documents/\` already contain useful context.
+3. Build the best current understanding from:
+   - root-level signals
    - existing workspace files
    - the user's message in this chat
-4. If the user mentions screenshots, notes, requirement docs, links, or file paths, tell them how those should be placed or referenced under \`.agentic/workspace/project/\`.
-5. If project materials are missing, say what would be useful to add, but continue with what you can understand already.
-6. Ask only blocking clarification questions.
-7. End with:
+4. If project context is thin, ask a small number of targeted questions about:
+   - project goal
+   - target users
+   - scope or modules
+   - available references such as screenshots, docs, notes, links, or file paths
+   - desired outputs
+5. If enough context exists, give a short summary of the project and the most important missing pieces.
+6. If the user mentions screenshots, notes, requirement docs, links, or file paths, tell them how those should be placed or referenced under \`.agentic/workspace/project/\`.
+7. After the first useful exchange, create or update:
+   - \`.agentic/workspace/memory/project-overview.md\`
+   - \`.agentic/workspace/memory/requirements.md\`
+   - \`.agentic/workspace/memory/open-questions.md\`
+8. Write drafts even from partial information.
+9. Clearly separate confirmed facts, assumptions, and open questions.
+10. Do not route immediately just because the workspace is empty. Do useful synthesis or ask focused questions first.
+11. Once you have done useful intake work, end with:
 
 \`Next type @.agentic/commands/project-requirements.md\`
 `,".agentic/context.md":`# Context
@@ -129,17 +112,20 @@ Read:
 
 ## Your job
 
-1. Collect the project explanation, goals, users, scope, screenshots, references, and known deliverables.
-2. Accept either:
+1. Refine and deepen the initial understanding created during \`@.agentic/init.md\`.
+2. Collect any missing project explanation, goals, users, scope, screenshots, references, and known deliverables.
+3. Accept either:
    - existing file paths already shared by the user, or
    - fresh explanation typed in chat
-3. Tell the user where each input belongs under \`.agentic/workspace/project/\`.
-4. Create or update:
+4. Tell the user where each input belongs under \`.agentic/workspace/project/\`.
+5. Organize and normalize the project context so the later architecture and document steps can rely on it.
+6. Create or update:
    - \`.agentic/workspace/memory/project-overview.md\`
    - \`.agentic/workspace/memory/requirements.md\`
    - \`.agentic/workspace/memory/open-questions.md\`
-5. Capture assumptions separately from confirmed facts.
-6. Ask only for missing details that block useful requirement understanding.
+7. Capture assumptions separately from confirmed facts.
+8. Reduce open questions where possible before handing off to architecture.
+9. Ask only for missing details that block useful requirement understanding.
 
 When done, tell the user:
 
@@ -249,15 +235,34 @@ Read:
 
 When done, tell the user the v1 workflow is complete and list the generated files.
 `}),b=`${p}
-## Agentic Starter
+# Agentic Starter Guide
+
+This project includes a reusable AI working area under \`.agentic/\` for project kick-off and pre-delivery documentation.
+
+## What This Is For
+
+Use this scaffold when you want an AI tool to:
+
+- understand a project from files, notes, screenshots, and links
+- maintain structured project memory
+- generate markdown deliverables such as BRD, FRD, estimate, and proposal
+
+## Folder Purpose
 
-This project includes a reusable AI workflow scaffold under \`.agentic/\`.
+- \`.agentic/init.md\`: first prompt for every new AI session
+- \`.agentic/context.md\`: shared rules and folder map
+- \`.agentic/commands/\`: task-specific prompt files you reference in your AI tool
+- \`.agentic/workspace/project/\`: raw project inputs such as screenshots, notes, requirement docs, and reference links
+- \`.agentic/workspace/memory/\`: evolving AI-maintained project understanding
+- \`.agentic/workspace/documents/\`: polished markdown deliverables
 
-Start every new AI session with:
+## How To Use It
 
-\`@.agentic/init.md\`
+1. Open your AI tool in this project.
+2. Start every new session with \`@.agentic/init.md\`.
+3. Follow the command sequence below.
 
-Command flow:
+## V1 Command Sequence
 
 1. \`@.agentic/init.md\`
 2. \`@.agentic/commands/project-requirements.md\`
@@ -266,16 +271,18 @@ Command flow:
 5. \`@.agentic/commands/create-frd.md\`
 6. \`@.agentic/commands/create-estimate.md\`
 7. \`@.agentic/commands/create-proposal.md\`
-${l}
-`,A=`${u}
+
+If your tool does not support \`@file\` references, paste the contents of the prompt file into a new chat manually.
+${u}
+`,q=`${l}
 .agentic/workspace/documents/
 .agentic/workspace/memory/
 ${g}
-`,R=[".agentic",".agentic/commands",".agentic/workspace",".agentic/workspace/project",".agentic/workspace/memory",".agentic/workspace/documents"];async function q(e){await Promise.all(v.map(t=>k(i.join(e,t),{recursive:!0,force:!0})))}async function E(e){await Promise.all(R.map(t=>f(i.join(e,t),{recursive:!0})))}async function I(e){let t=x();await Promise.all(Object.entries(t).map(([r,n])=>c(i.join(e,r),n,"utf8")))}function T(e,t,r,n){let a=new RegExp(`${m(r)}[\\s\\S]*?${m(n)}\\n?`,"m");if(a.test(e))return e.replace(a,`${t}
-`);let o=e.endsWith(`
+`,R=[".agentic",".agentic/commands",".agentic/workspace",".agentic/workspace/project",".agentic/workspace/memory",".agentic/workspace/documents"];async function A(e){await Promise.all(x.map(t=>k(i.join(e,t),{recursive:!0,force:!0})))}async function E(e){await Promise.all(R.map(t=>f(i.join(e,t),{recursive:!0})))}async function I(e){let t=v();await Promise.all(Object.entries(t).map(([o,n])=>c(i.join(e,o),n,"utf8")))}function T(e,t,o,n){let s=new RegExp(`${m(o)}[\\s\\S]*?${m(n)}\\n?`,"m");if(s.test(e))return e.replace(s,`${t}
+`);let r=e.endsWith(`
 `)?`
 `:`
 
-`;return`${e}${o}${t}
-`}function m(e){return e.replace(/[.*+?^${}()|[\]\\]/g,"\\$&")}async function d(e,t,r,n,a){let o=i.join(e,t);try{let s=await h(o,"utf8"),w=T(s,r,n,a);await c(o,w,"utf8")}catch(s){if(s.code!=="ENOENT")throw s;await c(o,`${r}
-`,"utf8")}}async function S(){let e=process.cwd();await q(e),await E(e),await I(e),await d(e,y,b,p,l),await d(e,j,A,u,g),console.log(""),console.log("create-agentic-starter: scaffold created successfully."),console.log(`Project folder: ${e}`),console.log(""),console.log("Next steps:"),console.log("1. Open your AI tool in this folder."),console.log("2. Start a new chat with @.agentic/init.md"),console.log("")}S().catch(e=>{let t=e instanceof Error?e.message:String(e);console.error(`create-agentic-starter: failed to scaffold project. ${t}`),process.exitCode=1});
+`;return`${e}${r}${t}
+`}function m(e){return e.replace(/[.*+?^${}()|[\]\\]/g,"\\$&")}async function d(e,t,o,n,s){let r=i.join(e,t);try{let a=await w(r,"utf8"),h=T(a,o,n,s);await c(r,h,"utf8")}catch(a){if(a.code!=="ENOENT")throw a;await c(r,`${o}
+`,"utf8")}}async function N(){let e=process.cwd();await A(e),await E(e),await I(e),await d(e,y,b,p,u),await d(e,j,q,l,g),console.log(""),console.log("create-agentic-starter: scaffold created successfully."),console.log(`Project folder: ${e}`),console.log(""),console.log("Next steps:"),console.log("1. Open your AI tool in this folder."),console.log("2. Start a new chat with @.agentic/init.md"),console.log("")}N().catch(e=>{let t=e instanceof Error?e.message:String(e);console.error(`create-agentic-starter: failed to scaffold project. ${t}`),process.exitCode=1});

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "create-agentic-starter",
-  "version": "0.1.3",
+  "version": "0.1.5",
   "description": "Scaffold a minimal AI-powered project delivery workflow inside the current folder.",
   "license": "MIT",
   "type": "module",