Hendri Martasari
Workflow

A grounded loop from intention to shipped outcomes.

This decision-driven flow keeps AI support anchored to code reality, product goals, and real constraints through repeated quality checkpoints.

01 / Product Intention

Define what the real problem and how the product should feel like before deciding how it should work.

user emotionproduct directionvisual tonetrust goal
02 / Structure

Translate the intent into flows, states, prompts, APIs, and technical boundaries.

UX flowedge casesAPI contractprompt logic
03 / Execution

Use AI tools to accelerate planning, coding, copywriting, and iteration.

Cursor plansSwiftUI screensbackend routesrefined copy
04 / Judgment

Review what to keep, simplify, rebuild, or ship.

product reviewrisk checkpolish passrelease decision

Product Intention / Problem

Start with the product need, friction, or goal that should drive the work.

Ask AI to Analyze the Situation

Use AI for structured analysis before deciding how to execute.

Extract Recommendation, Risks, Tradeoffs, and Hidden Failure Modes

Pull out concrete recommendations and likely failure points up front.

Reality Check the Advice

Validate the AI output against the actual environment before proceeding.

Decision

Does this match the actual codebase, product goal, and constraints?

No -> Loop back to: Ask AI to Analyze the Situation

Use Cursor Plan Mode

Switch to planning once the guidance is grounded in reality.

Ask Cursor to Inspect Codebase

Have Cursor map the real code state before implementation.

Ask Cursor to Break Work into Phases

Translate work into phased, reviewable chunks with clear intent.

Quality Control Checkpoint

Pause and verify plan quality before writing production code.

Decision

Is the plan grounded in real files, real state, and real APIs?

Ask Cursor to revise plan with evidence

Request a revision backed by concrete project evidence.

No -> Loop back to: Ask Cursor to Break Work into Phases

Build Phase by Phase

Implement incrementally while preserving the agreed plan boundaries.

Test in App / Simulator

Validate behavior in the real runtime context instead of assumptions.

Decision

Does result match intention?

Diagnose: bug, UX mismatch, state issue, API issue, or bad AI assumption

Classify the failure before rewriting instructions or code.

Rewrite instruction with tighter constraints

Constrain the prompt so the next pass stays aligned with intention.

No -> Loop back to: Use Cursor Plan Mode

Ship / Merge / Stabilize

Finalize the change safely once behavior matches the goal.

Document New Rule or Pattern

Capture durable guidance so future cycles are faster and safer.