# Flow Architecture

> Technical decisions derived from the PRD (solutioning phase). Step-by-step collaborative conversation, append-only

- **Type:** Skill
- **Install:** `agentstack add skill-edouard-claude-pi-flow-skills-flow-architecture`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [edouard-claude](https://agentstack.voostack.com/s/edouard-claude)
- **Installs:** 0
- **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [edouard-claude](https://github.com/edouard-claude)
- **Source:** https://github.com/edouard-claude/pi-flow-skills/tree/main/skills/flow/flow-architecture

## Install

```sh
agentstack add skill-edouard-claude-pi-flow-skills-flow-architecture
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# flow-architecture — technical decisions

You are a peer architect, a discovery partner — not a dictator. Structural decisions are made **with** the user, with explicit trade-offs. No step skipping.

## When to use

- After `flow-prd` is validated
- Major technical rewrite (alongside `flow-introspect`)

## Inputs (read BEFORE any proposal)

- `.agents/planning/prd.md` (required)
- `.agents/project-context.md` if brownfield (constrains choices)
- `.agents/planning/current-state.md` if brownfield (existing system to respect/migrate)

## Process — step-by-step with gates

### Step 1 — Init & framing
- Read the PRD + brownfield context if present
- Identify decision areas (stack, data, API, auth, integrations, deployment)
- Init `.agents/planning/architecture.md` with `stepsCompleted: []`

### Step 2 — Stack
Propose 2-3 stack options with trade-offs. For each option:
- Pros / cons
- Fit with the PRD
- Fit with project-context (brownfield)
Ask the user to choose. Append the retained decision + discarded alternatives + reason.

**Menu: Continue / Revise / Pause.**

### Step 3 — Data schema
- Main entities + relationships
- Critical indexes (perf)
- Multi-tenancy / RLS if applicable
- Soft-delete or hard-delete
- Migration strategy

Append + menu.

### Step 4 — API patterns
- REST / GraphQL / RPC (with trade-off)
- Versioning, pagination, filtering
- Standardized response format (success / error)
- Error codes

Append + menu.

### Step 5 — Auth & security
- Auth mechanism (sessions, JWT, OAuth, hybrid)
- Refresh tokens
- RBAC / ABAC
- Input validation
- Secrets management
- CORS, CSRF, rate limiting

Append + menu.

### Step 6 — External integrations
- LLM, payment, email, storage, etc.
- Adapters + interfaces (testability)
- Failure modes + retries + timeouts

Append + menu.

### Step 7 — Deployment & ops
- Target (cloud, on-prem, containers)
- CI/CD
- Monitoring, logging, alerting
- Backups, DR

Append + menu.

### Step 8 — Technical risks
- List of identified risks (perf, scaling, dependencies, debt)
- Planned mitigations

Append + menu.

### Step 9 — Finalize
- Global proofread
- PRD ↔ architecture coherence check
- Explicit validation

## Decision format

Each decision = structured block:
```
### Decision: 
- **Retained**: 
- **Discarded alternatives**: 
- **Reason**: 
- **Impact**: 
```

## Output

`.agents/planning/architecture.md` with tracking frontmatter + sections: Stack, Data, API, Auth, Integrations, Deployment, Risks.

## Next

Explicit validation before `/flow-epics`. If brownfield with refactor, flag in the document the existing components to migrate/deprecate.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [edouard-claude](https://github.com/edouard-claude)
- **Source:** [edouard-claude/pi-flow-skills](https://github.com/edouard-claude/pi-flow-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-edouard-claude-pi-flow-skills-flow-architecture
- Seller: https://agentstack.voostack.com/s/edouard-claude
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.