skills

My skills for coding agents (Claude Code, Codex, Cursor).

Install

npx skills@latest add javierscode/skills

Main workflow

From an idea to a pull request, one skill per step:

1. /define-scope — settle every decision with you, writing the glossary and the ADRs along the way.
2. /write-plan — cut that scope into vertical slices, one file per task.
3. /implement-plan — run the slices as reviewed subagents into a single pull request.

Skills

You invoke these

Typed as /name. Nothing else can reach them, so this list is what you have to remember.

• commit-work — cuts your working tree into one commit per intent, shows you the proposal, and commits once you approve.
• define-scope — interviews you one question at a time until every decision is settled, writing the glossary and the ADRs as it goes.
• implement-plan — runs a plan's slices as parallel subagents, each one reviewed before it merges, into a single pull request.
• write-plan — turns what you decided in the session into a plan of vertical slices, one file per task, written once you approve the breakdown.

The agent invokes these

Reached on its own when the task fits, or by the skills above. You can type them too.

• conventional-commits — where a commit is cut and how its message reads, so any skill that ends in commits writes them the same way.
• domain-modeling — sharpens your project's vocabulary while you talk, writing the glossary and the ADRs under it the moment they settle.
• interviewing — the design-tree interview itself, behind /define-scope. Type /interviewing for the interview without the docs.

How they compose

---
config:
  flowchart:
    rankSpacing: 70
    subGraphTitleMargin:
      top: 10
      bottom: 20
---
flowchart TB
  subgraph you["You invoke these"]
    direction LR
    define-scope(["/define-scope"])
    write-plan(["/write-plan"])
    implement-plan(["/implement-plan"])
    commit-work(["/commit-work"])
  end
  subgraph agent["The agent invokes these"]
    direction LR
    interviewing(["interviewing"])
    domain-modeling(["domain-modeling"])
    conventional-commits(["conventional-commits"])
  end

  define-scope --> interviewing
  define-scope --> domain-modeling
  implement-plan --> conventional-commits
  commit-work --> conventional-commits

  classDef user fill:#2563eb,stroke:#1d4ed8,color:#fff
  classDef model fill:#7c3aed,stroke:#6d28d9,color:#fff
  class define-scope,write-plan,implement-plan,commit-work user
  class interviewing,domain-modeling,conventional-commits model
  style you fill:transparent,stroke:#2563eb,stroke-dasharray:4 4
  style agent fill:transparent,stroke:#7c3aed,stroke-dasharray:4 4

Loading

License

MIT — see LICENSE.