agentleFS
RBraga01RBraga01/a-team/skills/smart-init/SKILL.md

smart-init

Skill24 starsChanged 24 days agoMITView on GitHub

Conversational onboarding for A Team. Detects ROADMAP.md, extracts context, generates INIT.md without requiring technical knowledge. Invoked automatically by the orchestrator when INIT.md is missing.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Check in this order:

    line 14
  2. Distinctive:

    Read the ROADMAP file and extract:

    line 25
  3. Distinctive:

    After extraction, ask only one question (the only thing a ROADMAP cannot tell you):

    line 37
  4. Distinctive:

    Ask these 5 questions, one at a time.

    line 52
  5. Distinctive:

    After Q5: generate INIT.md, show review gate, then offer to create ROADMAP.md:

    line 89
  6. Distinctive:

    Run silently after the user answers B.

    line 96
  7. Very specific:

    After generating INIT.md, display ONLY the ## O que entendi section and ask:

    line 118
  8. Distinctive:

    After 3 correction rounds without approval → ask user to edit INIT.md manually: "Não consegui perceber a correcção.

    line 124

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Smart Init — Conversational Onboarding

Trigger

This skill is invoked by the orchestrator when /orchestrate init is called and no INIT.md exists.

Detection Sequence

Check in this order:

  1. INIT.md exists → stop, use current init flow unchanged
  2. ROADMAP.md exists → Path A (extract from ROADMAP)
  3. ROADMAP_*.md exists (e.g. ROADMAP_icd10.md) → Path A (extract from that file)
  4. Neither exists → Path B (full 5-question interview)

Path A — ROADMAP Exists

Read the ROADMAP file and extract:

Look for in ROADMAPMaps to INIT.md field
Project name in H1 or titleProject name
Description paragraphProject overview
Stack table or technology mentionsLanguages & stack
"Princípios não-negociáveis" or "Non-negotiable" sectionImmutable rules
"Próximo" / "Next" / "Roadmap" itemsActive work context
Compliance mentions (GDPR, local-first, privacy, HIPAA)Compliance scope
"Feito" / "Done" sectionExisting coverage

After extraction, ask only one question (the only thing a ROADMAP cannot tell you):

"Que ferramentas de IA estás a usar para escrever código neste projecto?" → A) Claude Code → B) Codex CLI → C) Cursor → D) OpenCode → E) Várias — escolho mais do que uma

Then generate INIT.md (see Template section below) and show the review gate.


Path B — No ROADMAP (Full Interview)

Ask these 5 questions, one at a time. Use plain language — assume the user is non-technical.

Q1 (free text):

"O que queres construir? Descreve com as tuas palavras."

Q2 (multiple choice):

"Já existe código ou começas do zero?" → A) Começo do zero → B) Já existe código → C) Não tenho a certeza

If answer is B: run stack inference silently (see Stack Inference section).

Q3 (multiple choice):

"Que tipo de projecto é?" → A) App web (abre no browser) → B) App móvel (iPhone ou Android) → C) API ou serviço de backend → D) Análise de dados ou automação → E) Outro

Q4 (multiple choice):

"Em que dispositivo ou plataforma deve correr?" → A) Browser (qualquer dispositivo) → B) iPhone / iPad → C) Android → D) Desktop (Windows / Mac / Linux) → E) Servidor / cloud

Q5 (multiple choice):

"Que ferramentas de IA estás a usar para escrever código?" → A) Claude Code → B) Codex CLI → C) Cursor → D) OpenCode → E) Várias — escolho mais do que uma

After Q5: generate INIT.md, show review gate, then offer to create ROADMAP.md:

"Queres que eu crie um ROADMAP.md para este projecto com base no que descreveste? É útil para futuras sessões."


Stack Inference (Path B, answer B to Q2 only)

Run silently after the user answers B. Do NOT run for new projects (answer A or C).

git ls-files | sed 's/.*\.//' | sort | uniq -c | sort -rn | head -10

Extension → stack mapping:

  • .kt → Kotlin / Android (agents: kotlin-reviewer, compose-ui)
  • .swift → Swift / iOS (agents: swift-reviewer)
  • .py → Python (agents: python-reviewer)
  • .ts or .tsx → TypeScript / React (agents: code-reviewer)
  • .go → Go (agents: go-reviewer)
  • .rs → Rust (agents: rust-reviewer)
  • .dart → Flutter (agents: flutter-reviewer)
  • .java → Java / Android (agents: kotlin-reviewer as fallback)

Confidence threshold: if one extension accounts for >40% of tracked files, pre-fill Q3/Q4 and confirm with user. If ambiguous, ask normally.


Review Gate

After generating INIT.md, display ONLY the ## O que entendi section and ask:

"Está correcto? Falta alguma coisa?"

  • User says "ok" / "sim" / "yes" → run /orchestrate init automatically
  • User describes a correction → update the relevant INIT.md section, show ## O que entendi again
  • After 3 correction rounds without approval → ask user to edit INIT.md manually: "Não consegui perceber a correcção. Por favor edita o INIT.md directamente e diz 'ok' quando estiver pronto."

Codex Warning

Show this before the final approval:

"Última coisa: o Codex vai pedir para aprovares um script de segurança na próxima sessão. Clica em 'Trust' para continuar — é o script de estado da A Team."


INIT.md Template

Generate this file at the project root. Fill each field from the interview answers or ROADMAP extraction.

## O que entendi
[Plain language summary: what the project is, inferred stack, active AI platforms, agent count after init]

Se algo estiver errado, edita este ficheiro antes de continuar.

---

# INIT.md — [Project Name]

> Run `/orchestrate init` after reviewing this file.

## Project Overview

**Name:** [from Q1 or ROADMAP H1]
**Type:** [from Q3: web app / mobile app / API / data / other]
**Status:** [New project / Active development]
**Description:** [from Q1 free text or ROADMAP description]

## Languages & Stack

[Check all that apply — inferred from ROADMAP or stack scan]
- [ ] Kotlin
- [ ] Swift
- [ ] Python
- [ ] TypeScript / JavaScript
- [ ] Go
- [ ] Rust
- [ ] Flutter / Dart
- [ ] Other: ___

**Framework / UI:** [inferred or left blank]
**Database:** [inferred or left blank]
**Build system:** [inferred or left blank]

## Compliance Scope

[Extracted from ROADMAP compliance mentions, or left unchecked]
- [ ] GDPR
- [ ] Child privacy / COPPA
- [ ] Local-first / no external data
- [ ] HIPAA
- [ ] PCI-DSS

## Active AI Platforms

[From Q5 or Path A question]
- [ ] Claude Code
- [ ] Codex CLI
- [ ] Cursor
- [ ] OpenCode

## Non-Negotiable Rules

[Extracted from ROADMAP "princípios não-negociáveis" section, or left blank]

## Active Work / Next Steps

[Extracted from ROADMAP "Próximo" / "Next" section, or left blank]

## Agents to Prune

[Inferred from stack — list agents irrelevant to this project's languages/domain]

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.