Generic development workflow skills for Claude Code. Break down, shape, plan, implement, verify.
/plugin marketplace add teambrilliant/marketplace
/plugin install dev-skills@teambrilliant
| Skill | Invoke | What it does |
|---|---|---|
| loop-check | /dev-skills:loop-check |
Assess what's needed for autonomous feedback loops |
| product-primitives | /dev-skills:product-primitives |
System → fundamental primitives & building blocks |
| shaping-work | /dev-skills:shaping-work |
Rough idea → structured work definition |
| product-thinker | /dev-skills:product-thinker |
Product decisions, UX analysis, build-vs-buy |
| product-discovery | /dev-skills:product-discovery |
Validate ideas before committing to build |
| implementation-planning | /dev-skills:implementation-planning |
Ticket → technical implementation plan |
| implement-change | /dev-skills:implement-change |
Plan → working code, phase by phase |
| qa-test | /dev-skills:qa-test |
Browser-based QA verification via sub-agent |
| design-language | /dev-skills:design-language |
Capture + enforce a product's visual design language |
loop-check → primitives → discovery → shape → plan → implement → QA
-1 0 1 2 3 4 5
product-thinker
(0-2)
design-language
(2-5, cross-cutting)
-1. /dev-skills:loop-check — assess what's needed for autonomous feedback loops
0. /dev-skills:product-primitives — break system into deep, composable primitives
/dev-skills:product-discovery— validate whether an idea is worth building (4 risks, experiments, evidence gates)/dev-skills:shaping-work— define what to build (features, bugs, improvements)/dev-skills:implementation-planning— design how to build it/dev-skills:implement-change— build it phase by phase/dev-skills:qa-test— verify in browser
/dev-skills:product-thinker pairs with stages 0-2 — product decisions, UX analysis, build-vs-buy evaluation.
/dev-skills:design-language is cross-cutting across stages 2-5 — captures design inspiration into a living docs/design.md, checks implementations for drift, pairs with shape / plan / implement / QA.
Cross-cutting across shaping-work, implementation-planning, strategic-thinker, and implement-change. The reference doc at skills/implementation-planning/references/rollout-primitives.md defines two mechanisms and three decisions.
Two mechanisms:
- Expand-contract — evolve a shared contract (DB schema, public API, multi-consumer interface) safely by shipping additive changes first, migrating consumers, then removing the old path. The migration cadence IS the rollout.
- Feature flag — wraps a user-visible code path. Serves two purposes (either justifies a flag):
- Launch control — cohort, tier, geo, timing, %-rollout, A/B, dogfooding
- Reversibility — kill-switch for critical-path / uncertain / risky changes
Three decisions (every plan walks this tree):
- Contract test — is a shared contract changing? → expand-contract.
- Launch-strategy test — who should see this, and when? → flag (launch flag).
- Kill-switch test — if bad in prod, what would I do? → flag (risk flag).
Default is no flag, no expand-contract. Pick the lightest mechanism(s) that produce the launch control AND reversibility actually needed. One flag per feature, at the user-visible boundary. Bug fixes never get flags. Schemas can't be flagged — flag the consumer, expand-contract the schema.
Integration with tap-skills: the flag system itself (PostHog / LaunchDarkly / static config / DB-driven / none) is discovered from .tap/architecture.md, which tap-skills:tap-audit seeds during repo audit. Planners read it; they don't re-derive it on every plan.
Copy the template into any repo to get the full skill chain with smart routing:
cp "$(claude plugin path dev-skills)/references/CLAUDE.local.template.md" ./CLAUDE.local.mdThis gives you:
- Auto-routing based on existing artifacts (plans, research docs)
- Product thinking only when requirements are fuzzy
- Mandatory review gate before medium/large implementation
- QA verification after implementation
Then add project-specific sections (dev server, database, tools) below the workflow block.