CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

new-project-bootstrap

Comprehensive bootstrap flow for a brand-new project. Begins with a deep-dive architectural conversation, then runs a plan → implement → review chain, and finally writes the initial UML.md. Load this when the user starts pi in an empty or near-empty directory, or when explicitly asked to bootstrap a new project.

Quality

97%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

New Project Bootstrap

Use this when the current working directory is a fresh project (no UML.md, little or no source code). Drive the user through four phases in order. Do not skip phases. Do not start writing code before phase 3.

Phase 1 — Architectural deep dive (conversation)

Before any planning, hold a short structured conversation with the user. Ask the questions below one batch at a time, wait for answers, then move on. Stop early if the user says "just start".

Batch A — Product:

  1. What does this project do, in one sentence?
  2. Who uses it and how (CLI, library, web service, daemon, ...)?
  3. What is explicitly out of scope?

Batch B — Constraints: 4. Language / runtime / minimum versions? 5. Deployment target (local, container, edge, embedded, ...)? 6. Hard non-functional constraints (latency, offline, privacy, size)?

Batch C — Shape: 7. Expected lifetime: throwaway, prototype, long-lived/? 8. Solo or team? Open source? 9. Any existing systems it must integrate with?

Batch D — Architecture preferences: 10. Preferred architectural style (monolith, modular monolith, plugin, services)? 11. State / persistence needs? 12. Testing posture (unit-heavy, e2e-heavy, "I'll add tests later")?

Reflect back a 3–5 bullet summary of what you heard and ask the user to confirm before moving to phase 2.

Phase 2 — Plan

Write plan.md at the repo root following the planning-first skill's format if it is available; otherwise use this structure:

# Plan: <project name>
## Goal
## Architecture
  - chosen style + 1-paragraph rationale
  - top-level module list
## Tech choices
  - language, framework, key libs, build tool, test runner
## Milestone 1 — walking skeleton
## Milestone 2+ — features
## Risks
## Checklist

Show the plan to the user. Get explicit approval before phase 3.

Phase 3 — Implement (walking skeleton first)

Build Milestone 1 only: the smallest end-to-end thing that runs. Tick checklist items as you go. If reality contradicts the plan, stop and update the plan before continuing.

Phase 4 — Review and seal

When Milestone 1 runs:

  1. Self-review the diff: read every file you created, look for dead code, inconsistencies, missing error handling, missing tests for the happy path.
  2. Run the project (build + smoke test) and report results.
  3. Generate UML.md per the uml-maintenance skill. Use the architecture from phase 1 as the source of truth, not just what currently exists in code.
  4. Tell the user what is done, what is intentionally deferred, and what should happen next.

Guardrails

  • Do not invent product requirements the user did not state.
  • Do not pick exotic dependencies without flagging the choice.
  • Never skip phase 1 unless the user explicitly says "skip questions".
Repository

g23polar/dotfiles

Path

pi/agent/skills/new-project-bootstrap/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.