From mellow-engineering
Execute implementation with MCP tool integration. Usage: /workflows:work [optional plan reference]
npx claudepluginhub get-mellow/claude-plugins --plugin mellow-engineering[plan reference or task]workflows/# Implementation Execution Workflow You are starting the WORK phase of compound engineering. Your goal is to execute the implementation plan systematically with quality built in. ## Prerequisites Before starting work: 1. Ensure a plan exists (from `/workflows:plan`) or create one inline 2. Understand the platform(s) involved 3. Set up MCP tool sessions as needed ## Step 1: Load the Plan If a plan exists: If no plan exists, create a quick inline plan before proceeding. ## Step 2: Create Task List Use TodoWrite to track implementation progress: Mark tasks as `in_progress` as you w...
/workExecutes a plan file end-to-end with checks using the flow-work skill. Accepts plan file path, Beads ID, or title; prompts for input if none provided.
/workExecutes a development plan end-to-end with checks for <fn-N or idea>. Supports --branch (current|new|worktree) and --review (rp|export|none) options.
/workAuto-classifies tasks from descriptions, folder paths, or issue IDs and routes to workflow orchestrators; resumes existing tasks from .maister folders.
/workAuto-classifies tasks from descriptions, folder paths, or issue identifiers (GitHub/Jira/Azure DevOps) and routes to workflow orchestrators. Resumes existing tasks.
/workExecutes a work plan from a plan/spec/todo file path: clarifies requirements, sets up git branch or worktree, generates prioritized todos, and completes tasks maintaining quality.
/workExecutes epic tasks from blueprint using fresh Claude instances via relay script, with quality gate verification and configurable retries.
You are starting the WORK phase of compound engineering. Your goal is to execute the implementation plan systematically with quality built in.
Before starting work:
/workflows:plan) or create one inlineIf a plan exists:
Read: plans/[plan-name].md
If no plan exists, create a quick inline plan before proceeding.
Use TodoWrite to track implementation progress:
TodoWrite: Create tasks from the plan's implementation steps
Mark tasks as in_progress as you work on them and completed when done.
For every feature, follow this cycle:
1. Write a failing test
2. Write just enough code to make it pass
3. Refactor if needed
4. Repeat until feature is complete
This is non-negotiable. Tests come FIRST, then implementation.
1. Set up Xcode session:
mcp__xcodebuildmcp__session-set-defaults(
projectPath: "apps/mobile/Mellow.xcodeproj",
scheme: "Mellow",
simulatorName: "iPhone 16 Pro"
)
2. For each implementation task (TDD cycle):
a) Write a failing test FIRST:
@Test, #expect)mcp__xcodebuildmcp__test_simb) Write just enough code to pass:
Features/[FeatureName]/Screen/Features/[FeatureName]/Views/Features/[FeatureName]/ViewData/Features/[FeatureName]/ViewModel/mcp__xcodebuildmcp__test_simc) Use MellowDesignSystem:
// Colors
Color.mellow(.textPrimary)
Color.mellow(.appBackground)
Color.mellow(.accent)
// Typography
.mellowTitle(.m)
.mellowBody(.m)
.mellowLabel(.s)
// Spacing
MellowSpacing.m // 16pt
MellowSpacing.l // 24pt
// Buttons
.mellowPrimaryButton()
.mellowSecondaryButton()
d) Refactor if needed, then repeat the cycle.
3. Build verification:
mcp__xcodebuildmcp__build_sim
4. Run tests:
mcp__xcodebuildmcp__test_sim
5. Visual verification (if UI changes):
mcp__xcodebuildmcp__build_run_sim
mcp__xcodebuildmcp__screenshot
6. UI automation (if testing interactions):
mcp__xcodebuildmcp__describe_ui # Get element coordinates
mcp__xcodebuildmcp__tap(x: ..., y: ...)
1. For each implementation task (TDD cycle):
a) Write a failing test FIRST:
pnpm test --filter=@mellow/[app-name]b) Write just enough code to pass:
c) Example Fastify route:
import { z } from 'zod'
const bodySchema = z.object({
name: z.string().min(1),
})
export async function routes(fastify: FastifyInstance) {
fastify.post('/resource', {
schema: {
body: zodToJsonSchema(bodySchema),
},
}, async (request, reply) => {
const body = bodySchema.parse(request.body)
// Implementation
})
}
d) Refactor if needed, then repeat the cycle.
2. Type checking:
pnpm type-check --filter=@mellow/[app-name]
3. Linting:
pnpm lint --filter=@mellow/[app-name]
4. Run tests:
pnpm test --filter=@mellow/[app-name]
5. For web features, use Playwright:
mcp__plugin_playwright_playwright__browser_navigate(url: "http://localhost:3000/...")
mcp__plugin_playwright_playwright__browser_snapshot
1. Check current schema:
mcp__plugin_supabase_supabase__list_tables(project_id: "...", schemas: ["public"])
2. Create migration:
mcp__plugin_supabase_supabase__apply_migration(
project_id: "...",
name: "add_feature_table",
query: "CREATE TABLE ..."
)
3. Always include RLS policies:
-- Enable RLS
ALTER TABLE feature_table ENABLE ROW LEVEL SECURITY;
-- User isolation policy
CREATE POLICY "Users can only access own data"
ON feature_table
FOR ALL
USING (user_id = auth.uid());
4. Validate with advisors:
mcp__plugin_supabase_supabase__get_advisors(project_id: "...", type: "security")
mcp__plugin_supabase_supabase__get_advisors(project_id: "...", type: "performance")
5. Generate types:
pnpm types:generate
CRITICAL: This step MUST be run after EVERY implementation chunk, not just at the end.
Use the quality-gate skill for all verification:
| Changed Files | Skill Section |
|---|---|
*.ts, *.tsx | skills/quality-gate.md#typescript |
*.swift | skills/quality-gate.md#ios |
*.sql, supabase/** | skills/quality-gate.md#database |
Run all applicable checks IN PARALLEL.
See skills/quality-gate.md for:
CRITICAL: You may ONLY commit after ALL checks in Step 5 pass.
Verify Quality Gate passed:
Commit changes:
git add -A
git commit -m "feat: [description]
- [change 1]
- [change 2]"
Update task list: Mark all completed tasks as done.
Suggest next step:
Recommend running /workflows:review to validate the implementation.
quality-gate - Verification commands (Step 5)ios-ui-automation - iOS simulator interactionsweb-ui-automation - Playwright interactionsgit-workflow - Commit conventions (Step 6)