magarcia

Answer a question from primary sources and record every source checked under `research/`. Use when the user asks for a topic researched, for the evidence behind a claim, or when a decision needs facts you would otherwise guess at.

Test locale- and i18n-specific browser behavior with agent-browser by forcing a non-English JavaScript runtime locale through CDP. Use when reproducing or verifying bugs involving date/time markers, number or currency separators, Intl APIs, RTL behavior, or…

Monitor an open pull request through CI, review feedback, conflicts, draft-to-ready transitions, and merge readiness. Use after opening a PR, when the user says "babysit this PR" or "CI is failing," or as the final stage of $ship.

Deliver a feature or ticket end to end through requirements, exploration, test-driven implementation, documentation sync, verification, and a merged pull request. Use when the user asks to "build" or "implement" work end to end. Use $implement or $ship for…

Guide for building delightful, human-first command-line tools with proper command structure, output formatting, error handling, configuration, and extensibility. Use when creating CLI tools, adding subcommands, designing CLI output, handling errors in CLIs,…

Check design-system and UI consistency for diffs that touch user-facing surfaces. Use for visual changes, design checks, or as the UI stage of $ship, including small component, style, copy, and layout changes.

Find and fix repository documentation that a code change made false, including AGENTS.md, README files, docs, ADR references, command examples, APIs, configuration, and environment variables. Use after architectural or interface changes, renames, workflow…

Audit and improve the quality of AI-assisted, vibe-coded, branch, or work-in-progress code changes through eight specialist subagents covering duplication, type ownership, unused code, dependency cycles, weak typing, error handling, legacy paths, and…

Disciplined diagnose loop for hard bugs and perf regressions — build a feedback loop first, then reproduce → minimise → hypothesise → instrument → fix → regression-test. Auto-trigger when user reports a bug, says something is broken/throwing/failing/flaky,…

Interview the user relentlessly about a plan or design — one question at a time, with a recommended answer, walking down each branch of the decision tree. Auto-trigger before any non-trivial change in silkdown: new decoration node types, new widgets, public…

Grilling session that challenges a plan against silkdown's existing documentation — docs/architecture.md (the Live Preview model), docs/gotchas.md (failure modes with known answers), CLAUDE.md / AGENTS.md (working rules, verification expectations), and the…

React performance and correctness rules tuned for silkdown — a CodeMirror 6 wrapper library, not a Next.js app. Auto-trigger when editing the React adapter (packages/react/src/SilkdownEditor.tsx, packages/react/src/index.ts), reviewing or refactoring any…

Test-driven development with red-green-refactor (vertical tracer-bullet slices, not horizontal). Auto-trigger on any change in packages/core, packages/react, or e2e/ that adds or modifies behavior — new decoration types, widgets, parsers, public API surface,…

Conventions for writing tests in silkdown so they double as documentation and produce explicit, easy-to-pinpoint failure messages. Auto-trigger when creating or editing any file matching packages/*/test/**/*.test.{ts,tsx} or e2e/**/*.spec.ts, when adding new…

Rust-style strict TypeScript discipline for silkdown — branded types, exhaustive discriminated unions over enums, `assertNever` exhaustiveness checks, `as const` and `satisfies` for literal preservation, parse-don't-validate at boundaries, Result-style…

Step up one layer of abstraction and produce a map of relevant modules, callers, and data flow using silkdown's vocabulary (decorate/, widgets/, util/, plugin.ts, the Live Preview algorithm, the public API at packages/*/src/index.ts). Auto-trigger when the…

Guide for building delightful, human-first command-line tools with proper command structure, output formatting, error handling, configuration, and extensibility. Use when creating CLI tools, adding subcommands, designing CLI output, handling errors in CLIs,…

Obsidian plugin debugging and troubleshooting guide. TRIGGER when: diagnosing plugin errors, investigating unexpected behavior, fixing runtime issues, analyzing console errors from Obsidian, or when the user mentions DevTools, console errors, plugin crashes,…

Obsidian plugin development guide — architecture, API patterns, and conventions. TRIGGER when: writing or editing Obsidian plugin source code (TypeScript files importing from "obsidian"), creating plugin components (settings tabs, modals, views, commands),…

Obsidian plugin distribution, publishing, and release management. TRIGGER when: preparing a plugin release, publishing to the Obsidian community plugin list, setting up GitHub Actions for automated releases, creating manifest.json or versions.json, submitting…

Obsidian plugin testing patterns and strategies. TRIGGER when: writing tests for Obsidian plugins, setting up test infrastructure (vitest/jest), mocking the Obsidian API, writing unit tests for plugin logic, or when the user asks about testing strategies for…

Guides writing engaging technical blog posts that developers actually read. Triggers on: "write a blog post", "blog post about", "technical article", "write about [topic]", "review my blog post", "improve this article", "technical writing", "structure my…

Use when user says "translate articles", "translate blog posts", "translate missing articles", "add Spanish translations", "add Catalan translations", or "translate untranslated posts".

Remove AI writing patterns from prose. Use when drafting, editing, or reviewing text to eliminate predictable AI tells.

Use when user says "improve my writing", "make this clearer", "edit for clarity", "copyedit", "proofread", "review my prose", or asks for help with documentation, README files, commit messages, error messages, UI text, comments, or any prose humans will read.…

Author tools for the personal lean-tools CLI toolkit (stored in ~/.claude/tools/ globally or in a project's .claude/tools/ for repo-specific use). Use when the user asks to create a tool or script, when you notice yourself running the same multi-step command…

Write idiomatic, high-quality Rust code following community best practices. Use when writing, reviewing, or refactoring Rust code (.rs files, Cargo.toml), converting code to Rust, designing error types with Result/Option, implementing traits, or when…