# Init Project

> Inspect an existing repository, prepare an empty project, or map a migration before implementation begins. Use when an agent first enters an unfamiliar project or when a new project needs its stack, platform, test commands, documentation entrypoint, project memory, and next route established. Preserve existing agent configuration and do not scaffold a stack without confirmed intent.

- Skill: `ashermahonin/init-project` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add ashermahonin/init-project`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ashermahonin/init-project/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: ashermahonin (https://skillmd.com/u/ashermahonin)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/ashermahonin/init-project

---


# Init Project

## Purpose

Establish the project context before implementation. Inspect an existing repository, scaffold an empty one from an agreed intent, or combine both for a migration. Finish with a technical profile, project memory, and the next route.

## Inputs

1. Read `references/init-decision-tree.md`.
2. Inspect the directory shape and probe the stack markers listed in the reference. Avoid reading the whole repository before the likely entrypoints are known.
3. Choose fingerprint, scaffold, or hybrid mode.
4. If the user passed a free-text intent, route to `intake-coordinator` for a 30-second clarification before scaffolding.

## Decision process

1. Build the technical fingerprint: languages, frameworks, platforms, package manager, test and build commands, deployment target, CI, secret handling, observability, and canonical documentation entrypoint.
2. In scaffold mode, compare viable stacks against the confirmed intent. Verify current framework and platform behavior before selecting one.
3. In hybrid mode, map the existing boundaries before proposing the new slice or migration.
4. Preserve existing `AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, `.agents/`, `.claude/`, `.codex/`, `.gemini/`, and `.github/` configuration. Propose a merge when an overlay is needed; never replace it silently.
5. Seed project memory only when the package skeleton is available and the chosen destination does not conflict with an existing vault. Otherwise provide the minimal note set to create.
6. Record the mode, fingerprint, unknowns, next route, and the first command or prompt that continues the work.

## Decision boundaries

- Use Context7 MCP whenever the detected or proposed stack involves an external library, framework, platform, API, or CLI whose current behavior shapes the scaffold.
- Keep stack choices and rejected alternatives tied to the project's actual constraints.
- Never overwrite an existing agent instruction file, client directory, or project vault.
- Do not scaffold when product intent, target platform, or data boundary is still materially ambiguous.

## Decision record

- Stack fingerprint report (language, framework, platform, runtime, CI, secrets, observability)
- Project-mode classification: fingerprint, scaffold, or hybrid
- Proposed or merged agent overlay when the project needs one
- Seeded project memory when a destination was agreed
- Recommended skill chain and the first user prompt to send
- Open assumptions and stop conditions

## Ready when

- Never start `service-implementation` directly from `init-project`.
- Never scaffold a stack the user did not confirm.
- Never write secrets, tokens, or machine-specific absolute paths into seeded files.
- Never replace existing project artifacts silently.
- End with one concrete next action.

## Handoff

Hand off with: detected mode, stack fingerprint, seeded artifacts, chosen entrypoint skill, open clarifying questions, and the single next skill to invoke.

## References

- `references/init-decision-tree.md`: decision tree for choosing fingerprint, scaffold, or hybrid mode and the matching skill chain.