Claude Skills by Dev-Toolbelt

CLAUDE.md devteam section — opt-out rules and auto-routing triggers.

Backlog — epics, sprints, tasks, DoD, dependencies, estimates.

Code comments — none by default; type annotations and AAA markers.

Conventional Commits — types, scopes, breaking changes, layered.

Remote environment credentials (staging, production) — access and read-only enforcement.

Detect branch, worktree session, and staged changes for devteam.

Diataxis — tutorials, how-to, reference, explanation types.

Discovery before implementation — HARD-GATE, YAGNI, incremental.

Maintains docs/ — surgical patches, no duplicate content.

Templates for docs/ — project, tech, architecture, standards.

Automatic per-feature lessons-learned promotion at session close, scoped mirror of /learn.

Pre-delivery checklist for frontend tasks — run it before declaring any task complete.

Git — branching, commits, PR workflow, merge, release tagging.

Load when the user asks to push — watch GitHub Actions and auto-fix failures until green.

Incident response — severity, communication, mitigation, post-mortem.

Quiz-first rule — use AskUserQuestion for any user-facing choice, never plain text.

Detect and migrate a project from dev-team-agents v1 to v2 (multi-provider, symlinks).

Emit the agent's tier, model, and effort as a table before any other action.

Dev Team Agents notifications — format, types, conditions, rotation.

Pure-markdown output rules for plans, reports, reviews, and any structured agent output.

Plan-before-execute — structured plan, approval, replanning.

PR review — correctness, security, tests, performance, style.

Foundational rule — context, coexistence, Plan Mode, language, credentials.

Pre-release checklist for dev-team-agents — semver bump, tagging, rollback.

Registry of mandatory reuse/standardization rules and the review/lint gates that enforce them.

Routes PR review by diff — delegates to backend/frontend specialist.

Code reviewer shared rules — context, SonarQube, commit style.

Reviewer mindset — production-survival bias and critical questions.

How to write an operational runbook — incident response and maintenance procedures.

Run only tests covering the touched code; full suite only on explicit user request.

Setup health check — categories, commands, auto-fix actions.

Setup-assistant scan — commands, discovery table, skill/MCP checks.

Conditional agent spawn — path patterns and text triggers.

Spec layer between overview.md and sprints — testable criteria, auto contract gate.

Identify a project's technology stack from file signals — canonical signal-to-stack table.

Token optimization — efficient reads, bash ops, model selection.

User preferences — language and settings from preferences.json.

Periodic status-table check-ins while background sub-agents work, gated by credentials.json.

Git worktree per task — <worktree_path>/<context>/<title>, in .worktrees.

Creates and updates skills per agentskills.io spec.

Contract testing — Pact, schema validation, breaking change prevention.

Testing an SPA against a separate API — MSW handlers, async state coverage, factories.

Testing React hooks and Vue composables in isolation — renderHook, withSetup, async flows.

Load testing — smoke/load/stress/soak/spike profiles, SLO thresholds, tooling.

Mutation testing — code mutations to measure test suite quality.

Snapshot testing — when to use, maintenance, CI integration.

Test pyramid — unit, integration, E2E patterns and pitfall guidance.

Testing strategy — what, when, how; coverage planning.

Visual regression — screenshot diffing, Playwright/Chromatic, CI.

Ant Design — enterprise React UI with forms, tables; official MCP.