agentic

Complete development workflow from planning to GitHub PR using PARA methodology and RLM for large codebases. Use when implementing a feature, fixing a bug, refactoring, or taking work from plan to GitHub PR.

Analyze and improve performance: profile, find bottlenecks, optimize, and instrument code with observability for diagnosing performance issues (profiling, bottleneck tracing). Use when the user asks about performance, slow code, bottlenecks, profiling, optimization, or adding performance-specific observability.

Design and implement tests (unit, integration, e2e, accessibility, i18n), improve coverage, and run test suites. Use when the user asks to add tests, write tests, improve coverage, run tests, test accessibility, test translations, or implement test strategy.

Configure and fix CI/CD pipelines: build, test, lint, deploy, database migrations, environment config, observability setup (log shipping, metric collection, alert configuration), and releases. Use when the user asks about CI, pipeline, GitHub Actions, deploy, fix the build, environment variables, monitoring setup, or release process.

Debug failures systematically: reproduce, hypothesize, bisect, and fix. Use when the user reports a bug, asks why something fails, or wants to find the root cause.

Write commit messages, changelogs, release notes, and manage versioning following Conventional Commits with required scope. Use when the user asks for a commit message, changelog, release notes, versioning, tagging, or how to format commits.

Configure MCP (Model Context Protocol) servers for Claude or Cursor. Use when the user runs /setup, wants to add Atlassian, Datadog, or Playwright MCP, or set up API keys for MCP connections.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

Implement code using TDD (Test-Driven Development). Write tests first, then code. Use for features, bug fixes, and refactoring with Red-Green-Refactor cycle.

Structured workflow for AI-assisted development (Plan → Review → Execute → Summarize → Archive). Use when the user invokes /plan, /execute, /summarize, /archive, /check, /help, /init, /status, or when adding features, fixing non-trivial bugs, refactoring, or doing any task that will result in git changes. Skip for read-only queries, explanations, or quick fixes.

Review code and design for security vulnerabilities and insecure patterns. Use when the user asks for a security review, security audit, find vulnerabilities, or when assessing auth, crypto, input handling, or exposure of sensitive data.

Write technical specifications (tech specs): goals, scope, architecture, components, APIs, data models, database migrations, feature flags, risks, and rollout. Use when the user asks for a tech spec, technical design, architecture doc, design document, or when planning database changes or rollout strategies.

Review code for correctness, readability, maintainability, accessibility, i18n, and alignment with project conventions. Use when the user asks for a code review, review this PR, review my code, or when examining diffs, pull requests, or patches.

Upgrade and manage dependencies: resolve conflicts, update lockfiles, and check compatibility. Use when the user asks to upgrade deps, update packages, fix dependency conflict, or update lockfile.

Write and update documentation: README, API docs, ADRs, inline docs, and runbooks. Use when the user asks to document this, update README, write API docs, write ADR, or add inline documentation.

Refactor code safely: extract, rename, simplify, and reorganize without changing behavior. Use when the user asks to refactor, extract function, rename, simplify, or clean up this code.

Process large codebases (100+ files) using the Recursive Language Model pattern. Treats code as an external environment, using parallel background agents to map-reduce complex tasks without context rot.