review
Review software designs and implementations for user experience, simplicity, and maintainability; apply fixes when requested.
Create or update a concise PRD and implementation checklist for a feature or scope change.
$ npx -y skills add golbin/agent-skills --skill prd --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/prdContext preview
The summary Claude sees to decide when to auto-load this skill.
Create or update a concise PRD and implementation checklist for a feature or scope change.
name: prd description: Create or update a concise PRD and implementation checklist for a feature or scope change.
Keep the user's outcome, target design, and remaining work in one document. Planning-only requests end with the PRD; continue implementation when already requested.
Read the project evidence needed to identify the problem, affected users, and observable success conditions. Ask about missing decisions that materially affect scope or correctness; record reasonable assumptions for other details.
Choose the simplest design that meets the need, correcting the foundation when necessary. Follow Unix philosophy: focused, composable parts, explicit interfaces, and policy and side effects at boundaries. APIs and internal tools also serve users, including developers and operators.
Include compatibility, recovery, retries, or migration for concrete user, data, contract, or failure risks. Keep temporary mechanisms removable with a retirement condition. Avoid speculative features and abstractions.
Update an existing PRD when available. Otherwise follow the requested location or project convention, defaulting to `tasks/prd-<feature>.md`.
Include:
Use only the sections the task needs. Split out linked files when detailed evidence or independent work warrants it. Validation items should identify the behavior or contract to prove and a suitable check allowed by project rules.
Preserve user edits and completed work. Update the plan as evidence changes; record deferred requirements and distinguish implemented work from verified work. Mark items complete only with supporting evidence. Repeat checks only after a relevant change, failure, or unresolved concern.
Remove unsupported assumptions, unnecessary scope, and repetition. The PRD is ready when another contributor can execute it and judge success without guessing material decisions. Report its path and decisions still requiring the user.
Reusable agent skills for Codex and other Agent Skills compatible tools. Skills are packaged under skills// so this repository can hold multiple installable skills.