ManceRayder42

research-notebook

Deep research on any topic via the NotebookLM CLI. Use whenever the user requests research on a topic, in any project, in any language. Triggers on phrases like "research this", "deep dive on", "do research on", "dig into", "find out about", "investigate", "look into", "research X for me". Creates or reuses a NotebookLM notebook, ingests 5–10 fresh sources, asks structured questions, and returns a cited synthesis. For multi-faceted topics, signals the main thread to fan out parallel sub-agents (web + hub + qmd) alongside.

done

End-of-session wrap-up — saves a session note to the hub, updates CONTEXT.md, and commits hub changes. Use this skill whenever the user types `/done`, `/wrap`, says "wrap up the session" / "save the session", or signals the conversation is ending and they want it documented. Also use when the user is closing a coding/research session and wants the work captured before they walk away — even if they don't say "/done" explicitly. Do NOT auto-trigger; wait for an explicit closing signal from the user.

playwright-cli

Disk-based Playwright QA scan for a running dev server or deployed site — screenshots, console-error capture, page-error capture, and same-origin link checks in one process with one compact report. Use this instead of an interactive Playwright-MCP navigate/snapshot loop whenever you're verifying a KNOWN site you (or a build pipeline) control — confirming a change renders, checking a build for console errors, screenshotting routes at multiple viewports, or catching broken internal links. Not for driving an unfamiliar third-party page (a competitor site, a social profile, a SaaS UI you don't control the markup of) or anything needing live interaction (click, fill, drag) — that's still Playwright MCP.

voice

Turn spoken audio into text, and text into a spoken reply, via ElevenLabs (needs an ELEVENLABS_API_KEY). Use when the user attaches a voice message or audio file and wants it transcribed, or asks for a reply as speech/audio instead of text (including sending a voice note over the Telegram bridge). Transcription degrades to a local, free fallback when no key is configured — never blocks on a missing key, just says less accurately; speaking a reply has no local fallback and fails loudly instead.

learn

Use when the user shares a conclusion, note, or takeaway from something they're studying — a book, course, or article — and wants it captured into their personal learning archive. Trigger phrases include "what I learned today", "notes from the book", "conclusions from chapter X", plus the explicit `/learn` command. Also use when the user pastes raw notes/highlights from a book and asks to summarize them. Do NOT use for general reading recommendations, for ideas/brainstorms (that's a separate ideas-capture flow), or for daily reflective journaling with no book/material attached.

media-gen

Photoreal image + video generation and upscaling via fal.ai (pay-per-use, one API key, always-current models: Kling, Seedance, FLUX, nano-banana Pro, Topaz). Use for image-to-video ("animate this image", "turn this photo into a video"), photoreal video clips for websites/demos, hero video loops, upscaling images/video, or any gen-media need beyond simple local generation. Rewrites rough prompts into model-optimized ones, routes to the cheapest model that fits, saves outputs locally. NOT for code-rendered motion graphics or scroll-driven cinematic flythroughs — that needs dedicated tooling this plugin doesn't ship. For simple illustrations/UI imagery, a separate Gemini-based image skill (if the user has one installed) is usually cheaper — reach for media-gen when a named fal.ai model, image-to-video, photoreal quality, or upscaling is actually needed. ("nano-banana Pro" below is a fal.ai model slug, unrelated to any separately-installed nano-banana skill.)

memory-consolidation

Consolidation of recent Claude Code conversation logs into hub memory files. Extracts patterns, preferences, decisions. Updates patterns.md, conversation-log.md, CONTEXT.md. Archives entries older than 14 days. Adds wikilinks for hub-tracked projects. Audits frontmatter compliance. Surfaces stale projects (>30 days silent). Does NOT touch Claude Code's own built-in auto-memory. Invoke directly ("run memory consolidation") or wire up the opt-in scheduler in references/scheduling.md.

wiki-article

Write a wiki article capturing structure you learned about an active project in the user's hub. Use when you grepped or spawned Explore agents to understand a project and must record that knowledge before /done, or when the user asks to "write a wiki article", "document this in the wiki", or "add this to the wiki".

eli5

Explain any concept, codebase, or architecture like reader know nothing about it, as visual HTML artifact — big pictures, few words. Trigger on "/eli5 <topic>", "explain like I'm 5", "ELI5 this".

grill-me

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".