bassclef

Coding agents know how to code. bassclef helps them know how to build.

A set of files (rules, templates, and checks) that sit between you and Claude Code. Turns coding capacity into shipped software that holds together with product thinking, planning, standards, review, and learning.

What is here

Coding agents know how to code. bassclef helps them know how to build.

Tell bassclef what you're trying to accomplish. It reads your intent and picks the right disciplines from a catalog you can read. Then it carries them through planning, verification, and learning — inside Claude Code.

Concretely — bassclef is a set of skills you drop into Claude Code. You type /riff with a paragraph. Claude comes back with 2-3 clickable mocks. Pick one, run /launch, and Claude drafts a spec plus user stories. Run /build and Claude ships the plan as a pull request your reviewer can act on. The whole path fits inside 15 minutes on your first try.

AI gave you the production capacity of a team. Not the disciplines of one. The other 69 lite skills supply what the team used to distribute: personas, JTBD tasks, ADRs, spec, review, chronicles, standards. Add them as you notice yourself doing the same thing by hand two sessions in a row. bassclef does nothing on its own. No background updates. No writes to your repo unless you typed a skill.

Start with Get started — 15 minutes to a first PR. Or pick another section from the sidebar based on what you are trying to do.

About the catalog

bassclef ships 99 skills upstream across three sizes — lite, standard, and ultra. The catalog also ships rules, hooks, named thinkers (luminaries — Norman on affordances, Cooper on personas, Beck on test rhythm, and so on), agents, standards, and ADRs. Lite is the size you install by default:

  • 72 skills
  • 79 hooks
  • 76 rules
  • 43 luminaries
  • 4 agents
  • 90 standards
  • 6 ADRs
  • 439 entries total

The rest ship in standard (bigger size — more skills; some may need vendor credentials your harness does not provide) or ultra (adds Voyage embeddings for picking the right named thinker on a given task).

This site shows the skills with hand-written pages so far. More land as follow-on work. See the tier system for what each size gives you, or the live upstream catalog on GitHub for skills not yet documented here.

On this page

© 2025–2026 Sunjay Pandey·Privacy·Apache-2.0 code