AI Coding Spec Packet Before Code Generation
A copy-ready packet for giving AI coding tools a bounded task, acceptance criteria, file ownership, tests, and review evidence before code generation starts.
A focused archive of Spec Writing articles for spec-first teams.
A copy-ready packet for giving AI coding tools a bounded task, acceptance criteria, file ownership, tests, and review evidence before code generation starts.
Review AI-generated pull requests against acceptance criteria: inspect the diff, run evidence checks, and catch failures a quick skim misses.
Use test-evidence gates for AI-generated code: require meaningful tests before merge and catch hallucinated implementations before release.
Design API error taxonomies AI-generated clients can use, with stable codes, retry categories, and machine-readable details.
Follow a vague support ticket as it becomes a shippable technical spec using Spec Skills, guided questions, and review-ready output.
How to specify idempotency keys, deduplication windows, and state-machine transitions so retries and partial failures don't double-charge or corrupt data.
Write non-goals that stop scope creep, name deferred work, and give reviewers a clear boundary before implementation starts.
Design high-risk releases with stop-loss metrics, rollout windows, rollback types, and owners who can pause expansion.
How to write a webhook consumer spec: signature verification, replay protection, retry and backoff rules, ordering assumptions, and idempotent handler design.
Write edge cases QA can execute with concrete inputs, starting states, expected outputs, and release-blocking priority.
Avoid ten specification mistakes that hide decisions, blur acceptance criteria, skip failure paths, and push scope arguments into implementation.
Turn software requirements into testable specs with observable outputs, failure paths, evidence types, and QA-ready acceptance criteria.
Use this pre-coding review checklist to separate blockers from preferences, record accepted risks, and attach test evidence before implementation begins.