Workflows

18 step-by-step playbooks in .agents/workflows/ that guide AI coding agents through complex operations. Decision trees, not tutorials. Correct output on first attempt.

How They Work

AI reads a workflow before performing a task. Each workflow:

Loads context — specific SKILL.md sections + source files
Defines procedure — ordered steps with decision points
Enforces patterns — references actual conventions
Includes verification — output validation checklist

Inventory

Workflow	File	Purpose
Add Feature	`add-feature.md`	Full-stack: schema → action → UI → RLS → i18n
Add API Route	`add-api-route.md`	oRPC or Route Handler + auth + validation
Add UI Component	`add-ui-component.md`	Glassmorphism 2.0 + motion + a11y
Add Payment Provider	`add-payment-provider.md`	`PaymentProvider` interface implementation
Add Email Template	`add-email-template.md`	Email + i18n across 3 locales
Add Locale	`add-locale.md`	New language + 100% key parity
Add OAuth Provider	`add-oauth-provider.md`	Supabase OAuth + callback
Database Migration	`database-migration.md`	Schema → Drizzle push → RLS
Deploy	`deploy.md`	Vercel/Railway + env checklist
Debugging	`debugging.md`	Triage: logs → classify → fix → verify
Security Audit	`security-audit.md`	OWASP: auth → RLS → input → webhooks → headers
Testing	`testing.md`	Playwright E2E + Vitest unit
Remove Org Feature	`remove-org-feature.md`	Disable multi-tenancy safely
Remove Mock Data	`remove-mock-data.md`	Production cleanup
Add AI Provider	`add-ai-provider.md`	`AIProvider` interface implementation
Add Email Provider	`add-email-provider.md`	`EmailProvider` interface implementation
Manage Support	`manage-support-tickets.md`	Support tickets, AI escalation, super-admin workflow
Graphify	`graphify.md`	Knowledge graph orientation and dependency tracing

Usage

With AI Agents

Read .agents/workflows/add-feature.md, then build a [feature name] feature.

The workflow directs the agent to exact files, patterns, and verification steps.

With Slash Commands

/add-feature       → Full-stack feature
/add-api-route     → API endpoint
/add-ui-component  → Design system component
/database-migration → Schema change
/deploy            → Production deploy
/debugging         → Issue triage
/security-audit    → Security review
/testing           → Write tests
/add-ai-provider   → AI provider integration
/add-email-provider→ Email provider integration

Structure

Every workflow follows the Agent Skills open standard frontmatter:

---
name: add-feature
description: End-to-end workflow to add a new feature — database schema, server actions, UI components, i18n, analytics, and testing.
argument-hint: <feature-name>
---

## Pre-Work

Read: [SKILL.md sections + source files]

## Steps

1. [Action with file paths and patterns]
2. [Decision: if X → Y; if Z → W]
3. ...

## Verification

- [ ] TypeScript: npm run type-check
- [ ] Lint: npm run lint
- [ ] [Feature-specific checks]

Frontmatter Fields

Field	Required	Purpose
`name`	Yes	Unambiguous slash-command identifier (e.g., `add-feature`)
`description`	Yes	LLM-readable summary for model decision-making
`argument-hint`	Optional	Placeholder shown after `/command` (e.g., `<feature-name>`)
`disable-model-invocation`	Optional	Set `true` on destructive workflows (deploy, migration, removal) to prevent accidental agent auto-execution

Skill/Workflow Authoring Rules

Agent guidance follows the Agent Skills progressive-disclosure model:

Keep name lowercase kebab-case and aligned with the file or directory intent.
Keep description specific to trigger conditions; avoid broad "use for everything" descriptions.
Keep SKILL.md files focused and move heavy reference material into references/.
Reference supporting files with relative paths and tell the agent when to load them.
Update .agents/skills/*/references/ and this workflow inventory whenever architecture, payment behavior, testing counts, or route/component inventories change.
Keep payment workflows aligned with durable webhook event IDs, provider-resource idempotency, fail-closed account deletion, and same-origin redirect rules.

Relationship to Skills

Resource	Contains	Purpose
`AGENTS.md`	Root agent instructions	Context AI Agents
`CLAUDE.md`	Persistent instruction manual for Claude Code	Claude context
`.agents/skills/backend/`	Schema, patterns, security	Backend context
`.agents/skills/frontend/`	Design system, components, motion	Frontend context
`.agents/skills/i18n/`	Namespaces, routing, conventions	i18n context
`.agents/workflows/`	Step-by-step procedures	Task execution

Skills = "what the architecture looks like" | Workflows = "how to change it safely"