Replace ad-hoc AI prompts with a repeatable spec-driven workflow: from idea to verified spec.

Named after the tracer-bullet technique from The Pragmatic Programmer: Tracer + Kit.

Markdown skills, zero runtime deps. No build step, no project dependencies.

Why TracerKit?

AI assistants work best with small, well-scoped tasks — not sprawling layers or flat task lists. TracerKit structures every feature as tracer-bullet vertical slices: each phase cuts through every layer (schema → service → API → UI → tests) and is demoable on its own. Integration problems surface early, not at the end.

Four skills drive the workflow: define (/tk:prd), plan (/tk:plan), build (/tk:build), verify (/tk:check). The AI reads your specs directly, counts progress, and marks completed work done. Pure Markdown, zero runtime deps.

DEFINE          PLAN           BUILD            VERIFY
/tk:prd  ───▶  /tk:plan  ───▶  /tk:build  ───▶  /tk:check
  │               │               │                 │
  ▼               ▼               ▼                 ▼
PRD doc       Phased plan    Implement phase    Pass/fail
                              + feedback loops

Get Started

Marketplace install:

/plugin marketplace add helderberto/tracerkit
/plugin install tk@tracerkit-marketplace

CLI install:

npm install -g tracerkit
tracerkit init

Skills are installed to ~/.claude/skills/, available in every project. Safe to re-run.

tracerkit init .              # install to .claude/skills/ in current dir
tracerkit update .            # update project-scoped skills
tracerkit uninstall .         # remove project-scoped skills

Copy skills into .cursor/rules/, or use Notepads for on-demand context. See docs/cursor-setup.md.

Install as native skills for auto-discovery:

gemini skills install https://github.com/helderberto/tracerkit.git --path skills

Or add to GEMINI.md for persistent context. See docs/gemini-cli-setup.md.

Copy skills to .github/skills/ or add workflow rules to .github/copilot-instructions.md. See docs/copilot-setup.md.

Uses agent-driven skill execution via AGENTS.md. See docs/opencode-setup.md.

Skills are plain Markdown with YAML frontmatter — they work with any agent that accepts instruction files. Clone the repo and point your agent at the skills/ directory.

Workflow

You: /tk:prd add dark mode support
AI:  Written .tracerkit/prds/dark-mode-support.md
     Run `/tk:plan dark-mode-support` next?

You: /tk:plan dark-mode-support
AI:  Phase 1 — Theme visible end-to-end
     Phase 2 — User can toggle and persist preference
     Written .tracerkit/plans/dark-mode-support.md
     Run `/tk:check dark-mode-support` when ready?

You: /tk:build dark-mode-support
AI:  Phase 1 — Theme visible end-to-end (3 remaining)
     Implementing... all checks pass. Commit?

You: /tk:check dark-mode-support
AI:  Total: 7/7
     All phases complete — implementation verified.

Between sessions, /tk:brief shows active features and picks up where you left off:

You: /tk:brief
AI:  dark-mode-support (3/7)
       Phase 1 — Theme visible end-to-end: 3/3
     > Phase 2 — User can toggle and persist preference: 0/4

     **Focus → dark-mode-support**

See Examples for full walkthroughs.

Skills

| Skill | What it does | Output | | ------------------ | ------------------------------------------------------ | ------------------------------------ | | /tk:prd <idea> | Interview → codebase scan → structured PRD | .tracerkit/prds/<slug>.md | | /tk:plan <slug> | PRD → phased vertical slices, each demoable on its own | .tracerkit/plans/<slug>.md | | /tk:build <slug> | Implement next incomplete phase, run feedback loops | Code changes + checked items in plan | | /tk:brief | Feature dashboard with progress and suggested focus | Terminal only, no files | | /tk:check [slug] | Verify done-when checkboxes against codebase and tests | Checkboxes updated in plan |

Docs

| Document | Description | | -------------------------------------------- | ------------------------------------------------- | | Examples | Walk through end-to-end usage scenarios | | CLI Reference | Commands: init, update, uninstall | | Comparison | Compare TracerKit to Spec Kit, Kiro, and OpenSpec | | Cursor Setup | Use TracerKit skills in Cursor | | Gemini CLI Setup | Use TracerKit skills in Gemini CLI | | Copilot Setup | Use TracerKit skills in GitHub Copilot | | OpenCode Setup | Use TracerKit skills in OpenCode |

Contributing

1. Fork the repo and create a feature branch
2. Use TracerKit itself to plan your change (/tk:prd + /tk:plan)
3. Implement following the plan phases
4. npm run lint:fix && npm run test:run && npm run typecheck
5. Conventional Commits (enforced by commitlint)
6. Open a PR against main

Support

For support, please open a GitHub issue. We welcome bug reports, feature requests, and questions.

Acknowledgments

This project was born out of Claude Code for Real Engineers, a cohort by Matt Pocock. The hands-on approach to building real things with Claude Code sparked the idea for TracerKit. If you're serious about AI-assisted engineering, I can't recommend Matt's cohorts and content highly enough.

License

MIT License © helderberto