Recommended workflow

This is the story most teams want: same quality bar every feature, without re-explaining the stack.

The loop

Step-by-step

1. Write requirements in plain language

No ceremony — what users need, constraints, and out-of-scope. The agent already knows your framework and ORM from .ai/.

2. Let the agent propose a plan

Ask for a plan that references your modules, data model, and API surface. You are aiming for one coherent design, not a stream of edits.

3. Run plan-review before coding

Invoke the skill

In Cursor, use the skill UI or ask explicitly: “Follow the plan-review skill in .ai/skills/plan-review/.”
In Claude Code, open the skill from .claude/skills/plan-review/.
Other IDEs: use the path your adapter created (see Understanding the output).

Outcome: scope, architecture, data layer, API, and test strategy are sanity-checked before files multiply.

4. Implement

The agent should follow .ai/rules/* automatically — repository pattern, validation style, error handling, etc. If it drifts, point at the specific rule file.

5. Run code-review on the diff

Read-only review pass with a checklist (security, data layer, architecture, Git hygiene). Fix blocking items before asking for a human review.

6. Run qa

Builds, tests, and structured checks the skill describes — you get a clearer “done” signal than “looks fine in the editor.”

7. Run ship

Commit message conventions, push, and PR creation flow are encoded in the skill (often using gh where applicable).

Bonus skills (when they shine)

• document-release — after merge, sync README/CHANGELOG/API docs.
• retro — periodic look at churn and test health.
• db-migration-review — before risky migrations.
• api-contract-check — when response shapes or routes change.
• dependency-audit — upgrades and new packages.

After generation

1. Edit .ai/context/domain-map.md with real domains and folder notes.
2. Lock .ai/context/tech-stack.md to libraries you actually allow.
3. Use .ai/tracking/efficiency.md when the same mistake keeps recurring — that is a signal to patch a rule.

---

You are now at “mastery” for day-to-day use. To improve the generator itself or report issues, see Contributing & support.