sh5623

sh5623/self-improvement

Plugin Claude Code

When a convention bites you, fix the convention — an evidence-gated repair loop plus budgets and archiving that keep rule docs from growing without bound.

sh5623/self-improvement

Plugin Claude Code

When a convention bites you, fix the convention — an evidence-gated repair loop (detect → classify → verify → codify → propagate → record) plus budgets, demotion and archiving that keep rule docs from growing without bound. Stack-agnostic.

sh5623/self-improvement

Agent

The agent to delegate to when the classification, routing, or draft of a convention gap is unclear during self-improvement (si-improve). It runs the given gap plus evidence through the gates (duplication, contradiction, frozen, insufficient evidence), classifies it into the narrowest landing spot per the §routing…

sh5623/self-improvement

Hook

Runs when a session starts, running an inline shell check. From sh5623/self-improvement.

sh5623/self-improvement

Skill Claude CodeCodex

Use when a convention document has gone over budget (line count or block count), or when dead and duplicated clauses have piled up. Triggers include "the convention doc is too long", "clean up the conventions", "do an archive pass", the si-improve section 6 check reporting the log cap (15 blocks) exceeded, and an…

sh5623/self-improvement

Skill Claude CodeCodex

Use when a convention, document, or procedure was missing, wrong, or ambiguous and it cost you time during the work, and that gap looks likely to recur in other files, features, or tasks. Also use when the end-of-work check ("of what bit me, what will recur?") surfaces a gap, and on requests like "self-improvement"…

sh5623/self-improvement

Skill Claude CodeCodex

Use to bootstrap a project once, when it has no self-improvement data file (convention changelog and routing table). Triggers include "set up self-improvement", "si-init", and si-improve / si-archive / convention-smith failing to find the data file. Also re-run it right after a plugin update — it compares the data…

sh5623/fe-rail

Plugin Claude Code

Claude Code plugin marketplace for frontend projects — fe-rail.

sh5623/fe-rail

Plugin Claude Code

Harness dedicated to frontend projects. Supports Next.js App Router / Vite SPA + TypeScript, Tailwind v3/v4, shadcn/ui. Automates the spec→build→review→PR cycle.

sh5623/fe-rail

Instructions file

Repository instructions for fe-rail, a Claude Code plugin for React and TypeScript frontend projects. They describe its workflow from specification through building, review, and pull request, including testing, accessibility, and performance checks.

sh5623/fe-rail

Agent

Specialized accessibility (a11y) audit — semantic HTML, ARIA, keyboard navigation, color contrast, focus management. More precise than fe-reviewer's a11y axis. READ-ONLY. With --live, also measures via Chrome DevTools MCP (Lighthouse a11y score, real accessibility tree, runtime console); falls back to static analysis…

sh5623/fe-rail

Agent

Requirement gap analysis before/during the fe-spec stage. Systematically identifies unasked questions, undefined guardrails, scope creep, unverified assumptions, and edge cases. READ-ONLY.

sh5623/fe-rail

Agent

React/TS architecture advisory — component boundaries, data flow, state management, routing, module dependencies. Supports both Next.js App Router and Vite SPA (TanStack Router / React Router 7·8). READ-ONLY analysis only.

sh5623/fe-rail

Agent

Build/type/compile error-fixing specialist. Resolves errors with minimal changes only, no architectural changes. Auto-fix step triggered after the quality-gate hook detects errors.

sh5623/fe-rail

Agent

Decomposes PPT/planning decks (multi-slide, converted to PDF/PNG) into [policy & rules / screen mockups / screen flow / data & fields]. Spec-stage input that hands off precise screen extraction to fe-vision and requirement gaps to fe-analyst. READ-ONLY.

sh5623/fe-rail

Agent

Dedicated codebase exploration. Delegate when a search needs 3+ queries. Returns only summaries so raw code doesn't flow into the parent context. READ-ONLY.

sh5623/fe-rail

Agent

Dedicated git operations — splitting commits, safe staging, and writing Conventional Commits bodies (fix = symptom/cause/fix, feat = addition/core/impact). fe-pr-author owns PRs; this agent owns commits. Destructive commands forbidden.

sh5623/fe-rail

Agent

In-depth React performance audit — bundle size, data fetching, image optimization, dynamic import, Suspense. Supports both Next.js (RSC, next/image) and Vite SPA (bundle analysis, fetchpriority). More precise than fe-reviewer's performance axis. READ-ONLY. With --live, also measures via Chrome DevTools MCP…

sh5623/fe-rail

Agent

Synthesizes commits, diff, and spec to write the PR body (fix = symptom/cause/fix, feat = addition/core/impact blocks) and runs gh pr create. Delegate at the PR stage after fe-review passes — the main session only gets back the PR URL.

sh5623/fe-rail

Agent

Refactoring analysis specialist — evaluates 6 dimensions (complexity, duplication, naming, structure, patterns, type safety) with Before/After code and an impact × difficulty matrix. READ-ONLY.

sh5623/fe-rail

Agent

External docs/library research specialist — official Next.js/React/TS docs, GitHub Issues, Stack Overflow. Every piece of information must cite a source URL.

sh5623/fe-rail

Agent

4-axis frontend code review — type safety, performance, a11y, code quality. READ-ONLY, no direct edits. Delegate after fe-build completes, before the PR stage.

sh5623/fe-rail

Agent

Generates frontend test code — derives BDD scenarios and supports TDD Red-Green-Refactor. Pairs with fe-test-runner (generation vs. execution).

sh5623/fe-rail

Agent

Dedicated test execution. Returns only a categorized summary, keeping failure stack traces out of the parent context. Auto-detects jest/vitest/playwright.