# Flow Init

> Initializes, inspects, resumes, or migrates portable PWDEV Flow project state, routing requests to the appropriate workflow and creating or preserving configuration files.

- Skill: `pwdev-solucoes/flow-init` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add pwdev-solucoes/flow-init`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pwdev-solucoes/flow-init/raw
- Safety review: PASS (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity, Task Management
- Tags: Configuration, Migration, Project State, Pwdev Flow
- Author: pwdev-solucoes (https://skillmd.com/u/pwdev-solucoes)
- Updated: 2026-08-22
- Page: https://skillmd.com/skills/pwdev-solucoes/flow-init

---


# Initialize PWDEV Flow

Read [workflow](../../references/workflow.md), [artifacts](../../references/artifacts.md), [migration](../../references/migration.md), and [safety](../../references/safety.md) before changing project state.

## Procedure

1. Inspect the repository without reading secrets:
   - locate `AGENTS.md` and read applicable instructions;
   - treat `CLAUDE.md` as optional compatibility context when present;
   - inspect manifests, source layout, tests, and recent Git history;
   - never open `.env`, credentials, tokens, private keys, or certificates.
2. Check for `.planning/flow/config.json`, `.planning/flow/state.md`, and the legacy `.planning/config.json`.
3. Route the request as `inspect`, `initialize`, or `migrate`:
   - if Flow state exists, do not overwrite it; summarize the active workflow, last gate, artifacts, and next valid action;
   - if legacy config exists and Flow config does not, report a migration candidate instead of initializing over it;
   - use `migrate` only when the user asks to adopt the legacy workspace.
4. For `migrate`:
   - resolve [migrate_legacy.py](../../scripts/migrate_legacy.py) from this installed plugin and run `plan` with the active repository root;
   - present mappings, excluded field names, and conflicts without exposing excluded values;
   - stop for explicit approval before `apply`;
   - after approval, run `apply`, create missing Flow state/directories without overwriting existing paths, and inventory legacy artifacts in `.planning/flow/migration.md`;
   - do not copy or move legacy artifacts without a second approval for exact paths;
   - record `migrated` through the audit helper only when the migrated config enables audit.
5. If state does not exist and the user requested fresh initialization:
   - classify the repository as `greenfield` or `brownfield` from existing source and history;
   - create `.planning/flow/quick/` and `.planning/flow/reports/`;
   - create configuration and state using the exact schemas in [artifacts](../../references/artifacts.md);
   - set the artifact language from the user's current language unless explicitly configured otherwise;
   - set `auto_commit` to `false`.
6. Report created, migrated, or discovered paths and recommend one next action.

## Constraints

- Preserve existing configuration fields that this version does not recognize.
- Preserve `.planning/config.json` byte-for-byte during migration.
- Do not add project dependencies, hooks, branches, commits, or remote configuration.
- Do not create or rewrite `AGENTS.md` unless the user separately requests it.
- Keep all initialization writes under `.planning/flow/`.

## Output

Return `MODE`, `STATUS`, repository classification, state path, configuration path, `MIGRATION` when applicable, and `NEXT`. Link every real local artifact with an absolute path in the final response.