Ai-Stack

Useful AI things an engineer actually uses daily — starting with agent skills.

Skills here are plain SKILL.md files, so they work with Claude Code, Cursor, Codex, Copilot, and any other agent that reads the open skills format.

Install

All skills:

npx skills@latest add nazmulnahid-git/Ai-Stack

One skill:

npx skills@latest add nazmulnahid-git/Ai-Stack --skill prodcheck

This drops the skill into your project's .claude/skills/ (or your agent's equivalent). Invoke with /prodcheck, /prodtest, /mergemain, or just describe the task and let the agent pick.

As a Claude Code plugin

Installs once at user level and stays updatable via /plugin marketplace update:

/plugin marketplace add nazmulnahid-git/Ai-Stack
/plugin install ai-stack@ai-stack

Skills

Skill	What it does
prodcheck	Senior-engineer production readiness review. Verifies every finding against the real code before reporting — security, tenant data isolation, N+1 and DB problems, concurrency, failure modes. Asks before it fixes anything.
prodtest	Senior-QA pass on a new feature. Unit + functional/integration tests, then Playwright E2E with screenshots saved to a gitignored .prodtest/ folder. Asks upfront whether to fix bugs or only report them.
mergemain	Fetches main fresh, merges it into the current branch, and resolves conflicts by reading both sides in full context instead of picking a side. Escalates to you when intent is genuinely ambiguous.
shipit	Clean commits and PRs. Asks up front whether to use AI attribution or keep the commit fully human-authored, matches your repo's existing commit convention, and splits logical commits. Opens PRs as drafts, and updates the existing PR instead of opening a second one.

Layout

skills/
  <skill-name>/
    SKILL.md        # frontmatter: name + description, then the instructions
.claude-plugin/
  plugin.json       # plugin metadata (skills/ is auto-discovered)
  marketplace.json  # catalog entry, so /plugin install works too

Anything more than a SKILL.md (references, scripts, templates) goes inside that skill's own folder.

Adding a skill

1. mkdir skills/<name> and write a SKILL.md.
2. Frontmatter needs exactly two fields: name (lowercase, hyphens) and description. The description is what makes an agent decide to load the skill — write it as trigger conditions, not marketing.
3. Keep the body procedural: numbered steps, explicit rules, concrete output format. Say what the agent must not do.

License

MIT