flow-setup
/flow-setup
Konfiguruje Flow w repozytorium, przenosi starsze układy .agents, tworzy paczki wiedzy OKF i bazowe skille oraz agentów.
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 5
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
- SKILL.md zawiera komentarze HTML (5). Nie widać ich w podglądzie Markdown, a model je czyta.
Co robi i kiedy
Konfiguruje Flow w repozytorium, przenosi starsze układy .agents, tworzy paczki wiedzy OKF i bazowe skille oraz agentów.
Oryginalny opis (EN)
Use when initializing or aligning Flow in a repository, migrating legacy .agents/ layouts, scaffolding OKF v0.2 knowledge bundles, or generating baseline and domain .agents/skills/ and .agents/agents/.
Jak działa
- Trigger
- Workflow
- Guardrails
- Output
- Validation
Podgląd SKILL.md
---
name: flow-setup
description: "Use when initializing or aligning Flow in a repository, migrating legacy .agents/ layouts, scaffolding OKF v0.2 knowledge bundles, or generating baseline and domain .agents/skills/ and .agents/agents/."
disable-model-invocation: true
---
# Flow Setup
<!-- lifecycle-ownership: owner=flow-setup; operations=setup -->
## Trigger
Use for `/flow:setup`. Follow [`../flow/references/setup.md`](../flow/references/setup.md).
## Workflow
1. **Environment Detection, Brownfield Alignment, & Sharded Toolchain Discovery**:
- Inspect `.agents/setup-state.json` and `.agents/bundles/index.md`. If `.agents/` already exists, offer **Align & True-Up (Recommended)**, **Full Re-Setup**, or **Exit**.
- In alignment/brownfield mode, inventory legacy layouts (`.beads/`, `.agents/specs/`, `.agents/tracks/`, `.agents/transactions/`, flat `knowledge/patterns.md`, sibling PRD child specs with `parent_prd:` if desired, un-promoted `bundles/research/`, legacy `.agents/skills/flow-memory-keeper/`), ensure subdirectory `index.md` files and `log.md` carry zero YAML frontmatter and zero standalone `---` body rules, and migrate active work without data loss after user confirmation.
- Probe installed host binaries (`uv`, `python3`, `cargo`, `go`, `bun`, `pnpm`, `npm`, `make`, `just`, `pwsh`, `powershell`) alongside repository manifests. Fan out parallel read-only `@researcher` subsystem Scouts (`type_name="self"` with `agents/researcher.md` on Google Antigravity; `model="inherit"`).
2. **Scaffold Core Context Chapters & Rich OKF v0.2 Bundle (`.agents/bundles/`)**:
- Initialize `.agents/bundles/index.md` (`okf_version: "0.2"`) and frontmatter-free `.agents/bundles/log.md`, `product/index.md`, `knowledge/index.md`, `research/index.md`, and `specs/index.md`.
- Create Flow's core context chapters with bounded `<= 40`-line `<!-- truth: start -->` blocks and zero body `---` separators: `product/product.md`, `product/product-guidelines.md`, `product/tech-stack.md`, and `knowledge/workflow.md` (populated with verified host commands from [`../../templates/agent/workflow.md`](../../templates/agent/workflow.md)).
- Synthesize evidence-backed nested chapters across `knowledge/{architecture,patterns,domains,testing,gotchas,decisions}/` (via parallel domain synthesis subagents when `>= 2` domains exist) with `sources: [{ id, resource, title }]` + `[^id]:` footnotes and `verified: [{ by, at }]`.
3. **Seed Baseline Engineering Disciplines & Stack-Specific Project-Local Skills (`.agents/skills/`) and Subagents (`.agents/agents/`)**:
- Follow [`../multi-harness-plugins/SKILL.md`](../multi-harness-plugins/SKILL.md) and [`../flow-docs/references/agent-context.md`](../flow-docs/references/agent-context.md).
- Embed baseline engineering rules into the repository's knowledge and skill set: **Deep-Module & Seam Design**, **Active Domain Modeling & ADR Filter**, **Public-Seam Vertical-Slice TDD & Hypothesis-Driven Bug Diagnosis**, and **Present-Tense Cold-Reader Prose & Documentation Accuracy**.
- Generate consolidated domain skills (`.agents/skills/<domain>/SKILL.md` `<= 200` lines + `references/*.md`), domain review capability skills (`.agents/skills/review-<domain>/SKILL.md` + `references/checklist.md` + `references/false-positive-suppressions.md`), `.agents/skills/index.md`, and paired project-local subagents (`.agents/agents/<domain>-reviewer.md` and `.agents/agents/<domain>-specialist.md` with H1 `# ` headings, `mainAgent: false`, `subagent: true`, `model: inherit`, and `commandExecutionPolicy: sandbox`).
4. **Git Worktrees, Cross-Harness Permissions & `Stop` Hook Bootstrap, `AGENTS.md`, & `.agents/setup-state.json`**:
- Ask once via `structured-choice-v1` whether `.agents/` is **`shared`** (tracked in Git, Recommended; linked `git worktree` checkouts edit worktree `.agents/` directly) or **`local-only`** (stealth mode via `.git/info/exclude`; linked worktrees symlink `.agents` to the primary checkout and route all Flow Markdown edits there), and verify `.venv` / `node_modules` symlinks when running inside a linked worktree.
- Offer opt-in non-destructive harness policy, subagent permission allowlists (`.claude/settings.local.json` with backup, Codex linked-worktree `writable_roots` note, `.cursor/hooks.json`, `opencode.json`), update root `AGENTS.md` (preserving `<!-- project-customization: start -->` blocks), and write `.agents/setup-state.json` (`workflow_preferences.ignore_policy`, `workflow_preferences.model_routing: "auto"`).
## Guardrails
- Never copy Flow's global plugin skills (`flow`, `flow-*`, `okf`, `debloat`, `multi-harness-plugins`) or legacy `flow-memory-keeper` into `.agents/skills/`.
- Every materialized knowledge chapter must cite real repository paths (`path/to/file::Symbol`) in `sources:` and `[^id]` footnotes.
- Preserve existing custom content inside `<!-- project-customization: start -->` and `<!-- truth: start -->` blocks during alignment.
## Output
Return the initialized or aligned `.agents/bundles/` tree, generated `.agents/skills/` and `.agents/agents/` inventory, verified host commands, and next command (`/flow:spec` or `/flow:prd`).
## Validation
Confirm root `index.md` declares `okf_version: "0.2"`, all links in `index.md` and `.agents/skills/index.md` resolve on disk, every non-reserved `.md` file has valid OKF frontmatter, and `.agents/setup-state.json` records `setup_status: "complete"`.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · cofin/flow | 90 | Apache-2.0 |
| cofin/flow | 100 | Apache-2.0 |
Migawka bazy z 13.11.2026 · wersja paczki 6f916081ad67 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100flow
Prowadzi cykl życia pracy w repozytorium z .agents: planowanie od specyfikacji, śledzenie zadań w pakietach OKF, implementację TDD, synchronizację, przegląd i archiwizację. - 100open-knowledge-format
Definiuje paczki wiedzy w formacie Open Knowledge Format (OKF) v0.1 od Google: tworzy, waliduje i konwertuje wiedzę na ten standard, w tym dla wiki LLM i pakietów wiedzy agentów. - 90flow
Prowadzi repozytorium z katalogiem .agents przez cykl Flow: planowanie na podstawie specyfikacji, wdrażanie TDD, przegląd, zakończenie i archiwizację. - 100okf-maintain-knowledge
Tworzy i aktualizuje paczki wiedzy OKF o usługach, API, schematach i decyzjach, gdy zmienia się kod lub dokumentacja projektu. - 100okf
Tworzy, aktualizuje i wykorzystuje paczki wiedzy Open Knowledge Format (markdown z YAML) opisujące usługi, API, schematy, metryki, runbooki i decyzje. - 100okf
Tworzy i utrzymuje paczki dokumentacji w formacie Open Knowledge Format, czyta kontekst projektu z dokumentów i aktualizuje specyfikacje oraz ADR po sesji lub pull requeście.