golbin agent-skills
Two skills for Codex and Claude Code: prd keeps a living implementation checklist and review audits designs from the user's side
- 216 ★ on the repository
- Last commit
Minimal access
Why this level
- The skills carry instructions only; on a fix request the agent edits project files
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/golbin-agent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills carry instructions only; on a fix request the agent edits project files
This is third-party code. Review the repository files before installing.
What it does
The repository holds two skills and two templates. The `prd` skill creates or updates a short requirements document for a feature: problem, intended outcome, scope, success conditions, target design and a checklist ordered by dependencies. By default the file is `tasks/prd-<feature>.md`, and in an existing project the skill follows that project's convention. The document stays alive and the agent updates it as decisions change. The `review` skill checks designs and code in a fixed order: user experience, simplicity, maintainability. The agent walks through scenarios, interfaces, data ownership and tests across the whole workflow. A review-only request makes no edits. Fixes happen when someone asks for them. The `templates` folder holds `AGENTS.md` and `CLAUDE.md`: the first carries concise project instructions and the second points Claude Code to it. Both are copied into the project root with `cp`.
Who it is for. Developers who run tasks in Codex or Claude Code and want a short PRD with a checklist plus reviews that put the user scenario first.
Good fit when
- A new feature lands and the agent needs a document with goal, scope and an implementation checklist
- Requirements shifted mid-task and the checklist has to follow the decision
- You want a design or pull request judged by user experience as well as by code style
- A new project needs a starter `AGENTS.md` that Claude Code reads as well
Not a fit when
- You want a style or security checker: `review` weighs user experience and simplicity
- You need a formal spec with versions and sign-off: `prd` keeps a short working document
- The task is a one-line fix: a requirements document only slows it down
Example request
Use $prd to describe the CSV export feature and build an implementation checklist, then check the result with $reviewLimitations
- The skills are short and set a way of thinking, with no rule checklist. Results depend on the project context the agent reads.
- `install.sh` downloads an archive through `curl` and `tar`, so it needs access to raw.githubusercontent.com and bash.
- The repository holds only two skills, so other topics need other sets.
- The `agents/openai.yaml` files target Codex, and Claude Code uses only `SKILL.md`.
How to disable. Delete the `prd` and `review` folders from the agent skills directory, for example `~/.codex/skills` or `~/.claude/skills`. Copied `AGENTS.md` and `CLAUDE.md` are removed like ordinary project files.
FAQ
How do I install the golbin skills?
Run `npx skills add golbin/agent-skills --skill prd -g -a codex claude-code -y` and swap in `review` for the second skill. `gh skill install` and the installer inside Codex also work.
Where does `prd` save the document?
By default in `tasks/prd-<feature>.md`. If the project already has a PRD or another convention, the skill updates that file or follows the convention.
Does `review` edit code?
Only when asked. A review request returns findings without touching code, documents or requirement status, and a fix request allows changes within the agreed scope.
Do I need GitHub CLI?
No, it is one route. `gh skill install` needs GitHub CLI 2.90.0 or newer, while `npx skills` and `install.sh` work without it.
What do the templates contain?
`templates/AGENTS.md` holds concise project instructions for coding agents and `templates/CLAUDE.md` points Claude Code to `AGENTS.md`. Replace the placeholders with your own.
Discussion
No questions or reviews yet.
Related
SkillsEleven skills take an agent from an idea to verified code through a PRD and RFCs kept as files in the repository
A skills library for Claude Code, Cursor and Codex: team code review, TDD, stacked PRs and repository readiness scoring
14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
A skill for second-agent review of plans or code: Claude Code starts rounds, Codex reviews, and they exchange files
molyanov-ai-dev is a set of skills and agents for Claude Code and Codex: feature planning, project knowledge, code review.
Three skills for Claude Code, Cursor and OpenCode: an interview-driven PRD, task slices and a design doc before any code
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| golbin agent-skills (this card) | no | 216 | |
| AI PRD Workflow | no | 310 | a full PRD workflow, while this keeps a short checklist document |
| Simple Reviewer | no | 133 | a simplicity code review, while this puts the user scenario first |
README in short
The README lists two skills and two templates, then gives four install routes: `npx skills`, `gh skill`, the Codex installer and the `install.sh` script. It closes with examples of calling the skills by name and a short note on how they work together.
README badge
Original SKILL.md
--- name: prd description: Create or update a concise PRD and implementation checklist for a feature or scope change. --- # PRD 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. ## Define the target 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. ## Write the document Update an existing PRD when available. Otherwise follow the requested location or project convention, defaulting to `tasks/prd-<feature>.md`. Include: - Problem, intended outcome, scope, and success conditions. - Target design, supporting evidence, constraints, and open decisions. - A checklist ordered by dependencies, with implementation and validation outcomes. - Current status and consequential design or scope changes. 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. ## Maintain and finish Preserve user edits and completed work. Update the plan as evidence changes;