Skip to content

build-implement-code-guide.md

Latest commit

 

History

History
131 lines (98 loc) · 4.12 KB

build-implement-code-guide.md

File metadata and controls

131 lines (98 loc) · 4.12 KB

Build: Implement Code with Amp

Use this prompt‑first workflow to implement changes in small, reversible slices with strong validation and clear handoffs.

Note: Run Amp from your repo root (CLI) or open the workspace in your IDE; there’s no need to include repository paths in prompts.

Why this phase matters

  • Translates specs into reliable, maintainable code with tests.
  • Catches defects early via fast, incremental loops.
  • Preserves consistency by following style and contracts.
  • Reduces risk with feature flags and least‑privilege design.

Pre‑flight checklist

  • Scope and acceptance criteria are explicit; non‑goals clear.
  • Interfaces/contracts to touch are listed; flag strategy defined.
  • Local environment runs; build/test commands known; secrets managed.
  • Style/naming/commit conventions confirmed.
  • Safe branch: git switch -c feat/<slug>

1) Plan the change (scope, contracts)

Prompt (single message):

Outline the minimal stepwise plan for <change>. Define user impact, inputs/outputs, invariants,
updated contracts, flag strategy, and migration path. Return 3–6 small, reversible slices with
acceptance criteria and risks.

What you’ll get:

  • Sequenced plan with scoped diffs and tests per slice
  • Contracts and invariants to preserve
  • Risks and rollback notes

2) Read with intention (targeted paths)

Prompt (single message):

Given these targets <paths/types/functions>, summarize responsibilities, key call sites, side effects,
and invariants to preserve. Return only what’s essential for safe edits.

What you’ll get:

  • Focused reading outcomes for just‑enough context
  • Call‑site and side‑effect awareness
  • Hidden coupling and edge cases to watch

3) Implement (incremental diffs)

Prompt (single message):

Implement slice N: <small change>. Adhere to style/naming rules. Produce a focused diff and updated
tests/docstrings. Call out any necessary knock‑on changes and why.

What you’ll get:

  • Minimal diff with updated tests
  • Rationale for any adjacent edits
  • Clear next step for the next slice

4) Validate (build, diagnostics, tests)

Prompt (single message):

Run typecheck/lint/tests for touched areas. Report failures with smallest repro and propose fixes.
If behind flag <name>, verify both on/off behavior. Summarize pass/fail and next actions.

What you’ll get:

  • Targeted validation results and fixes
  • Confidence behind feature flags
  • Evidence for reviewer handoff

5) Risk & compliance (least privilege, data types)

Prompt (single message):

Review changes for least‑privilege access, sensitive data handling, input validation, and error
messaging. Propose stricter types or guards where appropriate and note any logging redactions.

What you’ll get:

  • Security/privacy checklist applied to the diff
  • Tighter types/guards and safer error handling
  • Logging guidance without secrets

Exit criteria and handoffs

  • Acceptance criteria pass locally; typecheck/lint/build/tests green
  • Tests added/updated for new behavior; brittle cases avoided
  • Feature guarded, off by default unless release‑ready; rollback documented
  • Review package prepared: summary, scope, risks, test evidence, rollout/rollback

Quick reference

Prompt templates

Plan: propose small, reversible slices with tests and risks.
Read: summarize responsibilities, call sites, side effects, invariants.
Implement: focused diff with updated tests and rationale.
Validate: typecheck/lint/tests; flag on/off; pass/fail and fixes.
Risk: least‑privilege, sensitive data, validation, error messaging; stricter types.

Developer‑run verification (optional)

# Build, typecheck, lint (adapt to your repo)
npm run build || pnpm build || yarn build
npm run typecheck || pnpm exec tsc -b || yarn tsc -b
npm run lint || pnpm lint || yarn lint

# Tests (focused first, then broader)
npm test -- --run --no-color || pnpm test -- --run --no-color || yarn test

Closing note

Keep changes small, typed, and reversible. Amp chooses tools automatically and returns synthesized results—not raw logs—so you can stay focused on outcomes.