scrapalot

Run the Scrapalot code improvement agent on scrapalot-chat (Python/FastAPI). Phases: 1=Code Style, 2=Architecture (analysis-only), 3=Docker Verify, 4=Audit. (No lint phase — the pre-commit hook runs ruff check + format on every commit.) Usage: /scrapalot:code-improver [1|2|3|4|style|architecture|docker|audit]

The rule for what a Scrapalot code comment may contain, and the procedure for cleaning up when it does not. Use when writing or reviewing comments and docstrings, or when sweeping a repo for commit/PR references, dates, owner attributions, corpus counts, or variable external data such as model names. Encodes what the corpus-wide cleanups learned, including the mistakes.

Turn finished research (competitive analyses, PRD backlog categories, partner PRDs) into shipped features. Owns three things the research skills do NOT: (1) the FILE-NAMING LAW — every research artifact has exactly one canonical path, and strays get renamed and relinked instead of quietly lost; (2) the INTERVIEW — the owner is questioned feature by feature in plain Croatian, with zero codenames, abbreviations or file paths, because they never read the research context; (3) the BUILD — worktree, feature branch, tests, PR, close-out. Invoked by /scrapalot:competitive-impl. Never invents work, never batches decisions.

Identify unused code in scrapalot-chat (Python/FastAPI). Runs parallel agents per directory area, produces a candidate list with confidence levels. Output is for human review — NOT auto-deletion.

Turn any outbound draft (Discord/DM/email/comment) into something a person will actually read. Removes AI style AND AI conversational behavior: retelling the reader their own message, flattery, process narration, scaffolding sentences, commitments the sender never approved, and replies that carry no new information. Step 0 picks the register — personal or analytical — because the rules differ. Applies to every outbound draft for every recipient; can also be invoked explicitly with /scrapalot:humanize-ai <text>.

Use this skill when adding or updating translations for Scrapalot UI. This includes adding new translation keys, translating UI text to multiple languages (English, Croatian, etc.), maintaining translation consistency, and ensuring all language files are kept in sync.

Use this skill when designing or building PUBLIC marketing/landing pages for Scrapalot (home, about, contact, buy-license, and src/components/landing). Pushes for art-directed, distinctive, motion-rich pages and breaks the generic-AI-UI defaults — adapted to Scrapalot's real stack (Framer Motion + existing landing primitives, NOT GSAP). For in-app product UI use scrapalot:ui-component-developer instead.

LangChain v1.x reference for the Scrapalot Python AI backend — APIs, migration guides and RAG/agentic patterns, backed by a scraped copy of the current docs. Use when writing or changing LangChain code in scrapalot-chat.

Rules for reading, writing, and validating Open Knowledge Format (OKF v0.1) bundles in Scrapalot — the corpus collection export and the second-brain export. Use when touching anything under src/main/service/knowledge_export/, when adding a new export surface, or when a user asks whether a bundle is portable/conformant.

Pydantic AI agent reference for the Scrapalot Python AI backend — creating, wiring and testing agents in the RAG pipeline, backed by a scraped copy of the current docs. Use when adding or changing a deep-research or RAG agent in scrapalot-chat.

Use this skill when developing new RAG (Retrieval Augmented Generation) strategies, orchestrators, or chunking methods for the Scrapalot-Chat project. This includes implementing new retrieval strategies, optimizing existing ones, adding orchestrators, or creating custom chunking approaches based on research papers or novel techniques.

Use this skill when optimizing public-facing pages (about, pricing, home, sign-up) for search engines. Focus on making Scrapalot discoverable for researchers in spirituality, consciousness studies, metaphysics, and non-mainstream science. Includes meta tags, structured data, semantic HTML, and content optimization for spiritual/consciousness research niches.

Use this skill when brainstorming, validating, or inventing unique features for Scrapalot focused on spirituality research, consciousness studies, metaphysics, and non-mainstream science. Provides deep analysis of competitive advantages, market positioning, and innovative feature ideas that differentiate Scrapalot in the spiritual/consciousness research niche.

Use this skill when analyzing technical debt in the Scrapalot-Chat Python/FastAPI codebase. Identifies code smells, RAG strategy inefficiencies, database design issues, missing tests, async/await anti-patterns, dependency problems, and creates comprehensive technical debt documentation.

Use this skill when building or modifying React components for the Scrapalot product UI (chat, notes, research, settings, knowledge, admin) using Radix UI primitives, CVA, TailwindCSS, Framer Motion and TypeScript. Encodes the real design system, the hard-won layout/Radix gotchas, and the Chrome-first verification protocol. NOT for public landing/marketing pages — use scrapalot:landing-design for those.