prodigeproject

Communicate efficiently without sacrificing clarity - natural, concise, actionable. Global skill loaded for every command and agent.

Use when reviewing or building UI components - ensures keyboard, screen-reader, and visual accessibility (WCAG 2.1 AA) so a11y is built in, not retrofitted.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Manages context caching to optimize token usage and cost by creating, incrementally updating, and invalidating caches while verifying integrity before reuse.

Clean code and engineering discipline: modularity, readability, sizing, naming, duplication, separation of concerns, plus the four behavioral principles - think before coding, simplicity first, surgical changes, and goal-driven execution. Global skill applied to all coding work.

Detects and resolves drift between code, documentation, and contextual knowledge, classifying each drift and recommending concrete sync actions to keep artifacts consistent.

Detects and quantifies technical, knowledge, architecture, documentation, and testing debt; rates severity and produces a prioritized paydown plan in time/risk/money terms.

Audits project dependencies for security vulnerabilities, license compliance, version conflicts, unused packages, and lockfile integrity. Owns supply-chain analysis (security-review defers here).

Generates a new AI prompt module with matching JSON schema, typed parser, and unit test. Use when adding a new AI capability (e.g., a new classifier, a new extraction prompt, or a new channel intent).

89 Claude Code workflow tips from Boris Cherny (creator of Claude Code) and the Claude Code team. PROACTIVE MODE: When the user is working on a task, check the CONTEXT MAP below and surface the most relevant tip BEFORE they ask. One tip at a time, short and actionable. Don't dump all tips — be a coach, not an encyclopedia. BROWSE MODE: When invoked with /boris, show the topic list and let the user pick. Use when: setting up Claude Code, optimizing workflows, running parallel sessions, configuring CLAUDE.md, using skills/commands, subagents, hooks, MCP integrations, worktrees, plan mode, verification, permissions, plugins, custom agents, sandboxing, keybindings, status lines, /simplify, /batch, /loop, /btw, /effort, /schedule, voice mode, remote control, auto mode, mobile app, session teleporting, Desktop app, Routines (scheduled/event-driven runs), /rewind, /compact vs /clear, auto-compact window, Opus 4.7 delegation model, full-context briefs, xhigh effort level, /fewer-permission-prompts, recaps, /focus, /g

Generates a new EmailProviderAdapter implementation skeleton that conforms to the contract in lib/providers/types.ts. Use when adding a new email provider (e.g., a new IMAP variant or a future provider).

Generates a new feature UI module skeleton following sekpriAI's component architecture. Use when adding a new screen or feature area to the app.

Generates RAG indexing and retrieval glue code. Use when adding a new source type to the RAG pipeline (e.g., a new attachment format) or when setting up the initial RAG infrastructure.

Generates a new Telegram/WhatsApp command handler that plugs into the channel intent system. Use when adding a new natural-language command that users can send via Telegram or the WhatsApp mock.

Generates test scaffolds for a feature module. Creates Vitest unit tests, integration tests, and Playwright E2E test skeletons following the testing spec (specs/008-testing-spec.md).

Pre-deploy checklist runner for Vercel deployment. Validates that all quality gates pass, env vars are configured, and the app is ready for production deployment.

Self-contained, zero-install token-saving system built on the four-layer token budget: say less prose (clarity-first concision), write less code (minimum correct solution), pull in less tool output (efficient tool use), and keep context lean (config audit + compaction-survival checkpoints). Ships a real CLI for per-session token tracking, honest savings measurement, and a quality guard so optimization never degrades code. ALWAYS active. Use when the user says "tokenme", "save tokens", "be efficient", "be concise", "minimal", or complains about cost, bloat, wordiness, or context filling up.

Create, redesign, critique, or plan audience-ready presentations and PowerPoint decks from a short prompt or source material. Use for portfolios, work/result reviews, proposals, sales decks, investor pitches, strategy or decision decks, research and technical talks, lessons, keynotes, reports, product launches, and reusable presentation templates. Infers user intent, audience, desired change, narrative, evidence needs, visual direction, delivery mode, and the best available slide runtime; produces editable PPTX, web slides, PDF, or a production-ready deck plan with rigorous content, visual, accessibility, and render QA.