Develop
Multi-account CLI companion for Claude Code that tracks usage limits, launches configured sessions, and applies skills & output styles.
npx -y skills add Tanq16/claudex --skill developAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Entry point for ANY coding work in a project that has these skills installed — implementing a feature, changing or refactoring code, fixing a bug, writing tests, scaffolding something new, or touching build/CI. Selects and loads the development skills that govern the task up front, holds the work to them while coding, and ends with a quick self-review that the skills were actually followed. Use this whenever you are about to develop anything.
SKILL.md
7.6 KB, as published. Nobody here has run it
Develop
The front door for development work: figure out the task, load the skills that govern it, follow them while coding, then verify you did.
The other skills carry the conventions; they only help if the right ones are in context before you write code and are still honored after. This skill makes that deterministic — it is a pre-step you run first, and a quick self-check you run last.
When to Use
Use this skill as the first step of any coding task in a project that has these skills installed:
- A new feature, a refactor, a bug fix, new or changed tests
- A new project or a new component scaffolded from scratch
- A build, CI/CD, Makefile, README, or config change
Run it before writing code — its whole job is to make sure the governing skills are loaded and applied, not to fix things afterward.
Not for:
- A pure question with no code change.
- A deliberate, thorough re-audit of an existing codebase — use
review-code(heavyweight, multi-agent).developonly self-reviews the diff you just wrote; see "Relationship to review-code" below.
Start here — required reading
develop has no references of its own — it routes you to the skills that do. Before writing any code:
- Read
go-foundations(../go-foundations/SKILL.md) in full for any Go work, ornode-foundations(../node-foundations/SKILL.md) for any Node work — always in scope. - Read, in full, the
SKILL.mdof every domain skill you select in Step 2 (Skill Map) — and honor each one's own Start here gate (its Always-tier references). Naming a skill is not reading it. - Delegating a sub-task? Brief the subagent with the same skills and their required references so it inherits the constraints.
Workflow
Step 1: Frame the task
State in one line what is being built or changed. Then classify:
- Project type — per the
go-foundationstaxonomy: CLI Only / Web Only / CLI + Web / Headless API Service / Library / Chrome Extension, or Node Web Only (a Node server process that serves a frontend). Infer it from the tree (go.mod,cmd/,internal/server/static/,manifest.json,package.jsonwith"type":"module"+public/, presence/absence ofutils/, …). - Work type — new project, feature, refactor, bug fix, tests, infra/CI, docs.
Step 2: Select and READ the governing skills
Use the Skill Map below to pick the skills this task touches. Then actually read each selected SKILL.md (and note which reference files will matter for this task) — do not just name them. go-foundations is always in scope for any Go work.
Pull in the skills the task actually touches — not all of them. When in doubt, load go-foundations plus the one domain skill that matches the files you are editing.
Step 3: State the rules in effect
Before writing code, emit a short checklist — the specific rules from the loaded skills that apply to this task, not a restatement of whole skills. A handful of concrete bullets, e.g.:
Rules in effect (CLI Only, new command):
- Comments: default none; keep one only if its *why* is load-bearing, judged on its own merit — why-not-what, one line, never restate code or embed scaffolding
- Output via utils printer (PrintInfo/PrintSuccess/...), never fmt.Println
- zerolog only behind --debug; human output through utils otherwise
- Tables via utils.PrintTable; honor --for-ai plain-text path
- New command file under cmd/, registered in root.go init()
A few rules are always in effect for any Go work regardless of task — first among them comment discipline (go-foundations → Comments and Code Style). Put these on the checklist every time: being cross-cutting rather than task-specific, they are the first to drop off the list and the first to decay mid-session. This keeps the rules in front of you while you code — it is the main defense against rules decaying out of attention mid-session.
Step 4: Do the work
Implement, holding to the Step 3 checklist. When the task spans several skills, keep each skill's rules in view for the part it governs (e.g. go-concurrency for the worker pool, go-cli for the command wiring).
For a large task, brief any sub-agents you spawn with the relevant skill(s) and their Step 3 rules so they inherit the same constraints.
Step 5: Quick self-review (inline)
After the change is done, do a fast pass over what you changed this session and check it against the rules you listed in Step 3. Fix gaps inline. This is a lightweight self-check, not a full audit:
- Scope it to the files you touched this session — not the whole project.
- Check the Step 3 checklist, not every rule in every skill.
- Fix small deviations directly; call out anything larger that needs a decision.
- No sub-agents, no domain orchestration — keep it quick.
Escalate to review-code instead when: you are re-engaging a project you have not touched in a while, the skills may have changed since the code was written, or you want a thorough multi-agent domain audit.
Skill Map
| Task touches… | Load |
|---|---|
| Any Go code | go-foundations (always) |
| Cobra commands, subcommands, flags, CLI output lifecycle, TUI | go-cli |
| HTTP servers, internal package architecture, storage, OAuth/auth | go-backend |
Embedded SPA frontend (embed.FS, Tailwind, Catppuccin, PWA) | go-frontend |
| Goroutines, concurrent pipelines, fan-out/fan-in, progress/resume | go-concurrency |
| Any Node code (Web Only: a server process that serves a frontend) | node-foundations (always, for Node) |
| Node HTTP/WebSocket server, routing, auth, JSON state | node-backend |
| Vanilla-JS SPA frontend (Catppuccin, vendored assets, WebSocket client) | node-frontend |
| Makefile, GitHub Actions, Docker, releases, versioning | project-ci-cd |
| README | project-readme |
| Chrome extension (manifest, popup, content/background scripts) | chrome-extension-basics |
For which skills a whole project type typically pulls in, defer to the go-foundations taxonomy and layout — it is the source of truth for project structure.
Relationship to review-code
These are deliberately different tools:
develop— lightweight, inline, during active coding. Confirms the change you just made follows the skills you loaded for it. Fast, scoped to your diff, no sub-agents.review-code— heavyweight, separate, multi-agent domain audit of an existing codebase. For re-engaging a project, or a deliberate compliance pass when skills may have changed.
Use develop every time you code. Reach for review-code when you specifically want the thorough audit.
Key Principles
- Skills first, code second — load and read the governing skills in Steps 1–2 before writing, not after.
- Skills are the single source of truth — only follow (and only self-flag) what a loaded skill defines. If no skill covers it, it is not a rule.
- Keep the checklist concrete — Step 3 lists the specific rules for this task, not whole skills restated.
- Always close with the self-review — Step 5 is cheap and catches the drift that creeps in mid-session.
- Don't over-load — pull in the skills the task actually touches, not the whole set.