intellectronica/agent-skills
44 skills89.4k installs12.3k starsGitHub
Install
npx skills add https://github.com/intellectronica/agent-skillsSkills in this repo
1Notion Apinotion-api is a reference agent skill that documents Notion API block types and JSON payload structures for developers building agents or automations that read and write Notion pages and databases.60.4kinstalls2Context7Context7 is an agent skill that fetches up-to-date documentation for software libraries, frameworks, and components by querying the Context7 API with curl. Developers use it when they need accurate API usage, code examples, or library behavior that may have changed since model training cutoff. The workflow is two steps: search for a library ID via the `/api/v2/libs/search` endpoint with `libraryName` and `query` parameters, then fetch topic-specific documentation from `/api/v2/context` using the returned `libraryId`, `query`, and optional `type=txt` for readable plain text. Examples in the skill cover React hooks, Next.js app router routing, and FastAPI dependency injection patterns. Tips recommend specific queries, jq filtering, URL-encoding spaces, and checking multiple search results when the first match is wrong. No API key is required for basic usage, though access is rate-limited on the public endpoints.9.2kinstalls3Youtube Transcriptyoutube-transcript extracts captions from YouTube videos using the youtube-transcript-api through a bundled uv script. Invoke it when a user provides a YouTube URL or 11-character video ID and asks for transcripts, subtitles, or captions. Supported URL formats include youtube.com watch links, youtu.be short links, embed URLs, and raw video IDs. Default output is plain text with one line per caption segment; add --timestamps for MM:SS or HH:MM:SS prefixed lines. The skill mandates never modifying the returned transcript text. For timestamp-free output, agents may reformat paragraphs without altering wording. Save to a user-specified file or default to VIDEO_ID-transcript.txt. The script fetches auto-generated or manually added captions, preferring manual captions when available and falling back to auto-generated tracks. Requires captions to be enabled on the target video. Run via uv run scripts/get_transcript.py with optional --timestamps flag for timed segments.3.1kinstalls4Nano Banana Pronano-banana-pro generates and edits images through Google's Nano Banana Pro API, implemented as Gemini 3 Pro Image, using a bundled Python script run with uv from the user's current working directory. Text-to-image mode passes a prompt and output filename; image-to-image editing adds --input-image without reading the source file into context first. Resolution flags accept 1K as default, 2K for medium output, and 4K for high resolution, mapping user phrases like 1080p or ultra to the correct API parameter. API keys resolve from --api-key or GEMINI_API_KEY environment variable with a hard error when missing. Filenames follow yyyy-mm-dd-hh-mm-ss-descriptive-name.png patterns with concise hyphenated descriptors. The skill instructs agents not to read generated images back and to report only the saved file path. Editing prompts cover add or remove elements, style changes, color adjustments, and background blur tasks on user-supplied image paths.2.9kinstalls5UltrathinkThe ultrathink skill is a minimal agent easter egg that displays colorful ANSI art spelling ultrathink when invoked. Triggers include the user saying ultrathink or invoking the slash ultrathink command. On activation it runs a single Bash printf command with multi-color escape sequences rendering block-letter ASCII art across three lines. The skill explicitly forbids adding any other output or commentary beyond the art display. There is no configuration, API integration, or follow-up workflow. It exists purely as a lightweight visual response for the ultrathink trigger phrase. Developers install it when they want a fun deterministic agent reaction tied to a specific keyword. Use only when the user explicitly requests ultrathink or the slash command, because the contract requires silent execution aside from the ANSI art itself with no explanatory text appended afterward.1.8kinstalls6Beautiful MermaidBeautiful Mermaid enables programmatic rendering of Mermaid diagrams to scalable SVG and high-resolution PNG formats. Supports flowcharts, sequence diagrams, state machines, class diagrams, and entity-relationship models with 13 customizable themes. Integrates with agent-browser for automated screenshot capture at 4K resolution. Includes syntax validation, edge label formatting, and special character handling. Outputs vector and raster assets to the current working directory with automatic cleanup of intermediary files.1.7kinstalls7Todoist ApiThis skill provides procedural guidance for interacting with Todoist using the td CLI tool. Developers use it to automate task, project, section, label, and comment management via natural language quick-add or explicit command flags. The tool supports JSON/NDJSON output for agent integration, filters for querying tasks by project/label/priority/due-date, and requires explicit user confirmation before destructive actions like delete or complete. Key workflows include listing tasks with filters, creating tasks with metadata, updating task properties, and managing projects with nesting and collaboration features.1.6kinstalls8Markdown ConverterThe markdown converter skill Convert documents and files to Markdown using markitdown. Use when converting PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xls), HTML, CSV, JSON, XML, images (with EXIF/OCR), audio (with transcription), ZIP archives, YouTube URLs, or EPubs to Markdown format for LLM processing or text analysis. Documentation covers workflows, commands, and guardrails agents should follow when users invoke this capability. Key documented areas include **Documents**: PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xls); **Web/Data**: HTML, CSV, JSON, XML; **Media**: Images (EXIF + OCR), Audio (EXIF + transcription); **Other**: ZIP (iterates contents), YouTube URLs, EPub. Reference commands include uvx markitdown input.pdf; uvx markitdown input.pdf -o output.md. Use when developers or agents need structured guidance for markdown converter tasks with evidence grounded in the bundled SKILL.md rather than generic advice.1.5kinstalls9Anki Connectanki-connect is an integration skill for developers who want coding agents to operate Anki through the AnkiConnect plugin instead of manual GUI clicks. The skill verifies Anki is running, waits for the AnkiConnect server at http://127.0.0.1:8765, then translates requests into JSON API calls executed with curl, jq, or equivalent shell tools. Developers use anki-connect to list or modify decks, notes, cards, note models, attached media, and sync operations directly from the terminal or agent chat. The workflow emphasizes safe request crafting and response interpretation so bulk card edits, deck imports, and media updates do not corrupt collections.719installs10Gog Cligog-cli documents gogcli installation, credential management, and account authorization for Google Desktop OAuth clients. Commands include brew install gogcli, gog auth credentials set with a client_secret JSON, gog auth add with comma-separated services like gmail,calendar,drive, plus gog auth list --check and gog auth status for verification. The skill stresses enabling only required Google Cloud APIs and checking live service names via gog auth services before prescribing scopes. Developers invoke gog-cli when diagnosing auth failures, picking credential modes, or wiring least-privilege agent access to Google Workspace user data.685installs11Nano Banana 2nano-banana-2 is an agent skill that wraps Google's Nano Banana 2 model (`gemini-3.1-flash-image-preview`) in a Python script runnable via `uv run`. Developers pass a text prompt to `generate_image.py` to create new images or edit existing ones, with explicit aspect-ratio control and output sizes from 512px through 4K. The skill installs under `.agents/skills/nano-banana-2/` and is triggered when a user asks to create, modify, or iterate on images inside Cursor or Claude Code. Reach for nano-banana-2 when you need quick visual drafts for UI concepts, app store screenshots, documentation diagrams, or marketing creatives without leaving the coding agent. It requires a Gemini API key and network access but keeps iteration inside the agent workflow instead of switching to a separate design tool.683installs12Promptifypromptify is a prompt-engineering skill in intellectronica/agent-skills that rewrites vague user requests into clear, specific, and complete prompts treated as specification language rather than casual natural language. The process reads full request context, removes ambiguity, and outputs prompts optimized for AI model consumption so Claude, Cursor, or Codex produce the intended result on the first pass. Developers reach for promptify when a task description is underspecified, when model outputs miss requirements, or when users explicitly say promptify or request prompt improvement. The skill triggers on promptify, promptify this, or explicit prompt-engineering requests.575installs13Tavilytavily is an intellectlectronica/agent-skills integration that teaches agents to call Tavily's REST API with curl for web search, page extraction, site mapping, crawling, and optional deep research. The skill returns structured, LLM-friendly summaries and chunked content suited to multi-step investigations when built-in browser tools are missing or when Tavily's output format is preferable. Developers reach for tavily when an agent task needs live documentation, competitor pages, API references, or site structure discovery without hand-rolling scrapers. The workflow emphasizes curl-based requests, parsing Tavily JSON responses, and chaining search with extraction or crawl steps across a research session. Use tavily inside Claude Code, Cursor, or similar agents whenever external web evidence must be fetched programmatically and fed back into reasoning loops with minimal formatting cleanup.570installs14Mgrep Code Searchmgrep-code-search is a semantic code exploration skill from intellectronica/agent-skills built around the mgrep CLI. It enables natural-language queries across code, text, PDFs, and images, complementing grep and ripgrep when pattern matching cannot surface intent or feature location. The skill recommends use when a repository has more than 30 non-gitignored files or nested directory structures that make manual path guessing slow. Developers and agents reach for mgrep-code-search to find where features live, understand unfamiliar modules, and explore large monorepos without writing precise regex first. It pairs traditional exact search with embedding-based retrieval for codebase onboarding and architecture discovery.559installs15Gpt Image 1 5gpt-image-1-5 is an intellectlectronica/agent-skills skill that lets coding agents generate or edit images with OpenAI's GPT Image 1.5 model through scripts/generate_image.py run via uv. Text-to-image generation uses the Responses API image_generation tool, while editing uses the Image API images.edit endpoint with optional --input-image and --mask for inpainting. Quality flags are low, medium, and high; size options are 1024x1024, 1024x1536, 1536x1024, and auto; background can be transparent, opaque, or auto on generation. The script reads OPENAI_API_KEY from the environment or --api-key and saves timestamped PNG files to the developer's current working directory. Developers reach for gpt-image-1-5 when asked to create, modify, change backgrounds, or inpaint regions in an existing image without manually opening an image editor.554installs16Copilot Sdkcopilot-sdk is an AI & Agent Building skill from intellectronica/agent-skills for embedding Microsoft and GitHub Copilot SDK capabilities into custom software. The skill covers SDK feature setup, authentication configuration, and tool wiring so autonomous agents can invoke Copilot-backed completions, chat, or action endpoints from application code. Developers reach for copilot-sdk when building internal dev tools, IDE extensions, or agent orchestrators that must integrate official Copilot APIs rather than raw OpenAI calls alone. It targets SDK-level integration work spanning auth, session management, and tool registration across multiple implementation steps.479installs17Here Be Githere-be-git is a Git & Pull Requests agent skill from intellectronica/agent-skills that runs a three-step workflow when a developer says "here be git", "init git", or wants version control in the current directory. Step 1 runs git init and confirms the repository. Step 2 optionally appends mandatory agent commit instructions to AGENTS.md—or creates that file—requiring commits after every task with no exceptions. Step 3 optionally generates a .gitignore from flavours like Node.js, Python, macOS, or IDE patterns, then commits each change atomically. If git is already initialized, the skill skips init and proceeds to agent instructions. here-be-git helps teams prevent agents from leaving uncommitted changes and establish repository hygiene before feature work begins.473installs18Ray So Code Snippetray-so-code-snippet is a skill from intellectronica/agent-skills that generates styled code-snippet images using ray.so and saves output to the working directory or a user-specified path. The workflow requires the developer or agent to supply source code directly or from a file selection, then collect all styling parameters—theme, language, padding, and background options—before rendering. Developers reach for ray-so-code-snippet when README examples, technical blog posts, slide decks, or social posts need screenshot-quality code blocks without opening a design tool. The skill automates fetching ray.so renders and persisting PNG or image files locally so documentation updates stay consistent and on-brand across repositories and talks.468installs19Lorem Ipsumlorem-ipsum is a frontend placeholder-copy skill from intellectonica/agent-skills that fills mockups, landing pages, and component stories with realistic text at specific lengths. Developers invoke it when Storybook stories, marketing sections, or UI components need believable filler so truncation, line breaks, and right-to-left rendering can be evaluated before copywriting finishes. The skill produces paragraph, headline, and label-length variants tuned to component constraints rather than generic lorem ipsum blocks. Reach for lorem-ipsum during design-system work, responsive breakpoint checks, and pre-launch page scaffolding where empty or one-word placeholders hide real layout bugs.455installs20Raindrop Apiraindrop-api is a Claude Code skill from intellectronica/agent-skills for comprehensive Raindrop.io REST API interaction at `https://api.raindrop.io/rest/v1/`. It resolves `RAINDROP_TOKEN` from environment or user input, verifies auth with `GET /user`, and covers CRUD for raindrops, collections, tags, highlights, filters, import, export, and backups. List endpoints paginate with `page` and `perpage` where max `perpage` is 50; the API rate limit is 120 requests per minute. Destructive DELETE and bulk operations require user confirmation. Export supports csv, html, and zip formats; OAuth access tokens expire after two weeks with refresh_token flow documented. Developers use raindrop-api when building research pipelines, share lists, or personal knowledge features that sync bookmarks, collections, and tags through Raindrop.io.447installs21Upstash Redis Kvupstash-redis-kv is a Claude Code skill from intellectronica/agent-skills for integrating Upstash Redis as a serverless key-value store in agents and HTTP APIs. It covers caching hot reads, session persistence, rate-limit counters, and short-lived agent state without provisioning or operating Redis clusters. Developers reach for upstash-redis-kv when deploying on Vercel, Cloudflare Workers, or other serverless runtimes that need durable KV with REST or HTTP-based Redis clients. The skill fits JWT session blacklists, API throttling, and cross-request agent memory that must survive cold starts.409installs22Monologue Notes Apimonologue-notes-api from intellectronica/agent-skills documents how agents integrate the Monologue notes API for programmatic capture and retrieval. It targets API, agent, and SaaS builds that need reliable note sync, auth, and CRUD patterns against Monologue during implementation.155installs23Anki Connectanki-connect is an agent skill that routes Anki-related requests through the AnkiConnect add-on API instead of manual deck editing. Developers use it to bulk-create flashcards from specs, import glossary terms from codebases, audit deck structure, or trigger sync after content pipelines run. The skill applies whenever a user wants to interact with Anki—including read-only inspection of models and cards or write operations that change notes and media. It assumes Anki is running locally with AnkiConnect listening on its default port. Prism catalogs it under Build → integrations because it is a task-specific bridge between your agent and a desktop app, similar to other local service plugins. Complexity is beginner-friendly for simple queries but intermediate when batch updates or custom note types are involved.0installs24Beautiful Mermaidbeautiful-mermaid is an agent skill that renders Mermaid diagram definitions into SVG and PNG using the Beautiful Mermaid library. Developers use it when they are writing documentation, planning system flows, or explaining data models and want visual output without leaving Claude Code, Cursor, or Codex. The skill is intentionally narrow: invoke it when the user asks to render a Mermaid diagram, not for general illustration or non-Mermaid formats. Because output is generated locally through the bundled plugin workflow, developers can iterate on diagram syntax in chat and drop the resulting assets into repos or landing pages. Pair it with planning and review skills when you need a diagram before implementation or after a design change.0installs25Context7context7 is an agent skill that routes documentation questions through the Context7 API so answers reflect current library releases, not memorized cutoff knowledge. developers hit it when implementing unfamiliar SDKs, double-checking parameter shapes, or pulling short examples for a specific API surface before committing code. The skill is appropriate whenever the task is “what does this framework do today?” rather than product positioning or deployment runbooks. It complements local repo search: Context7 resolves canonical package docs and examples you can paste into plans or tests. Multi-phase placement reflects repeated use during Build feature work, Ship review fixes, and Operate migrations when upstream APIs change. Keep network access scoped and prefer verifying critical security-sensitive APIs against official sources when stakes are high.0installs26Copilot Sdkcopilot-sdk is an agent skill for developers embedding GitHub Copilot programmatically instead of relying only on IDE completions. It walks through SDK setup across Node.js/TypeScript, Python, Go,.NET, and Java, then into the parts teams usually get wrong on the first pass: authentication, permission scopes, streaming events, and durable sessions. The skill also covers extension points—custom tools, custom agents, MCP servers, hooks, and Copilot skills—so you can wire your product’s domain logic into a Copilot host. Developers shipping internal agents or customer-facing assistants on Microsoft’s stack use it during implementation spikes and hardening, not as a marketing or SEO utility. Expect integration-heavy guidance rather than a single copy-paste template; you still bring your repo, identity provider choices, and deployment target.0installs27Gog Cligog-cli is an agent skill that teaches Claude Code, Cursor, and similar agents to run Google Workspace operations through the gog (gogcli) command-line tool. It targets developers who want repeatable Gmail, Calendar, Drive, and productivity-app workflows without maintaining bespoke OAuth scripts for every task. Invoke it when users name gog, need agent-safe Google automation, or want to search mail, manage events, edit Docs/Sheets, administer Workspace, or script Apps Script from the terminal. The skill maps user intent to the right gog subcommands, auth flows, and configuration so agents do not improvise dangerous or incomplete Google API calls. It matters on Prism because Workspace is where many solos store specs, customer lists, and launch assets—this skill makes that stack programmable from the same session where they ship code.0installs28Gpt Image 1 5gpt-image-1-5 is an agent skill that wires OpenAI’s GPT Image 1.5 into your coding workflow for developers who need screenshots, hero art, icons, or quick visual iterations without opening a separate design tool for every tweak. When you ask to generate an image from a prompt—or to modify one you already have—the skill routes the job through the model’s text-to-image and edit paths, including optional masks for partial updates. The documented flow emphasizes calling the skill with an input image parameter rather than loading the file into context first, which keeps token use predictable on large PNGs. It fits early UI exploration, app-store graphics, and content refreshes, but it depends on OpenAI image API access and acceptable usage policies. It is a task integration, not a local offline renderer.0installs29Here Be Githere-be-git is a lightweight agent skill for developers who want version control without a separate onboarding checklist. When you say you want Git initialized in the directory you are already working in, the skill runs a focused setup flow: initialize the repository, apply a.gitignore when appropriate, and optionally drop in commit instructions your Claude Code, Cursor, or Codex session can follow later. It is meant for the first minutes of a new or unpacked project, not for advanced branch policies or team governance. The value is speed and consistency—you get a traceable history and fewer accidental commits of build artifacts before you dive into features. Pair it with your normal review habits once code starts moving.0installs30Lorem Ipsumlorem-ipsum is a small agent skill that generates lorem ipsum and generic placeholder text on request. developers use it while mocking UIs, storybook states, landing page sections, or documentation templates where real copy is not ready yet. It supports several structures—plain paragraphs, headed sections, bullet or numbered lists, and flowing continuous text—and can return content for immediate use in a component or save it to a file. The skill fits the build phase when you want realistic text density without distracting copywriting. It keeps layout and typography decisions honest before validate/launch content work begins. Because triggers are explicit (user asks for placeholder or dummy text), it stays out of the way during research or production SEO. Install it when your agent frequently hand-writes filler sentences instead of consistent structured blocks.0installs31Markdown Convertermarkdown-converter is an agent skill that wraps markitdown to convert heterogeneous files into Markdown. Developers use it when raw inputs—client PDFs, slide decks, spreadsheets, scraped HTML, or archived ZIP bundles—block an agent from quoting, diffing, or chunking content. The skill is invoked explicitly when you need conversion before summarization, spec extraction, or RAG indexing, rather than as a background formatter. It fits developers shipping agent-first products, content pipelines, or internal tools that must treat documents as first-class context. Placement on Build → docs reflects the usual handoff: ingest → normalize to Markdown → feed planning, implementation, or support bots. It also appears in Idea and Validate when you analyze market PDFs or pricing spreadsheets without retyping them.0installs32Mgrep Code Searchmgrep-code-search is an agent skill that wraps mgrep semantic search for developers exploring sizable or deeply nested repositories. It is meant for moments when you know what you are looking for in plain language—authentication flow, retry logic, feature flags—but exact strings are unknown or scattered. The skill positions semantic retrieval alongside traditional grep and ripgrep so your coding agent can map unfamiliar modules, trace cross-cutting concerns, and validate assumptions before edits. Install it when your project crosses the documented threshold of more than thirty non-gitignored files or when folder depth makes linear search tedious. It supports Claude Code, Cursor, Codex, and similar agents that need procedural guidance for when to reach for mgrep instead of burning tokens on blind file walks. Complexity is intermediate because you still must phrase queries well and verify hits against source truth.0installs33Monologue Notes Apimonologue-notes-api documents how a coding agent should read and search a developer’s Monologue notes through the official REST API. founders who capture ideas, meeting notes, and specs in Monologue install this when they want Claude or Cursor to ground answers in their private notebook instead of hallucinating project history. The skill is procedural: set MONOLOGUE_API_KEY, avoid leaking the key in logs, paginate list endpoints, apply filters, and retrieve individual note bodies with predictable HTTP calls. It does not write or mutate notes—read-only access keeps blast radius small for automation. Use it while building features that reference prior decisions, validating scope against written notes, or growing content workflows that reuse captured research. Prism places it on Build → integrations because the work product is correct API usage, not a standalone doc generator.0installs34Nano Banana 2nano-banana-2 wraps Google’s Nano Banana 2 image API so agents can generate and revise visuals with controlled aspect ratio and resolution from half-K through 4K. developers use it when validating a product vibe on a landing page mock, drafting App Store or social creatives before hiring a designer, or swapping hero art during a launch week without opening a separate image tool chain. The skill targets speed and parameter control rather than a full design system workflow, so you still own brand prompts and review for policy compliance on generated content. It requires Google API access appropriate to Gemini image preview endpoints. Treat outputs as prototypes unless you run your own QA pass for text-in-image accuracy and trademark issues.0installs35Nano Banana Pronano-banana-pro is an agent skill that wires your coding agent to Google’s Nano Banana Pro (Gemini 3 Pro Image) API for both creating images from prompts and editing existing files. Developers use it when they need hero shots, social graphics, UI concept art, or quick variants without leaving the agent thread. Triggers match natural language asks to generate, create, edit, modify, change, alter, or update images, including when the user points at an on-disk image and wants background swaps or object replacements. The workflow favors passing --input-image directly rather than pre-reading binary assets, which keeps token use down and matches the skill’s documented guardrail. Resolution can be tuned for default 1K delivery or sharper 2K and 4K exports when you are shipping landing pages or store assets. It is a task integration, not a full creative pipeline—you still own brand consistency, licensing, and prompt safety review before publishing.0installs36Notion Apinotion-api is a procedural agent skill that teaches your assistant how to call the Notion REST API correctly when you want specs, roadmaps, or customer notes living in Notion to stay in sync with code. developers use it to list and query databases, append blocks to pages, update properties, manage comments, and handle deletes without guessing endpoint shapes or cursor pagination. The skill emphasizes authentication, available endpoints, pagination, and error handling so agents fail loudly instead of silently truncating results. It fits agents that can issue HTTP requests with your integration token rather than replacing the official Notion MCP. Invoke it during build when you automate onboarding docs, sync issue status from a repo, or prototype a lightweight CMS backed by Notion. Skip it if you only need one-off copy-paste from Notion without ongoing API automation.0installs37Promptifypromptify is a compact agent skill that performs prompt engineering on demand: you hand it a half-formed goal, and it expands it into a precise brief models can execute. Developers hit it constantly because agent quality is mostly prompt quality—whether you are ideating positioning copy, scoping an MVP, drafting tests, or writing runbooks. The skill listens for direct triggers such as “promptify” or explicit requests to improve a prompt, then adds role, context, constraints, output format, and success criteria without you manually repeating best practices. It does not replace domain skills like brainstorming or writing-plans; it makes those invocations start from a stronger user message. Complexity stays beginner-friendly: no toolchain beyond your agent session. Use it whenever a single chat message would otherwise produce generic or incomplete work product.0installs38Raindrop Apiraindrop-api teaches coding agents how to interact with Raindrop.io through its REST API using curl and jq, so developers can automate bookmark libraries instead of clicking through the web app. Use it when you want to pull competitor saves into a research doc, file launch links into collections, dedupe tags before a content sprint, or export backups as part of a personal knowledge pipeline. The skill spans authentication setup through advanced filters and import/export, making it a procedural integration reference rather than a hosted MCP server. You bring a Raindrop API token and network access; the agent follows the documented endpoints. It pairs well with research and writing workflows but stays anchored as an API integration skill for developers who already rely on Raindrop as their link inbox.0installs39Ray So Code Snippetray-so-code-snippet is an agent skill that produces polished code snippet images through ray.so so developers can share readable screenshots in threads, newsletters, and README files. You invoke it when you already have the code and want a visual asset without fiddling with fonts and padding in Figma or a manual ray.so session. It fits developers shipping tutorials, open-source repos, or launch announcements where one clear screenshot beats a pasted block. The skill emphasizes local save paths so assets land next to your repo or content draft. It is a narrow generator, not a general screenshot or IDE capture tool—supply the snippet and formatting intent and let the agent render and store the image.0installs40Tavilytavily is an agent skill that documents how to call Tavily’s REST API for search, content extraction, mapping, and crawling when your coding agent lacks a native browser or search tool, or when you want responses formatted for downstream LLM reasoning. Developers use it during idea research, validation spikes, and build-time doc lookups without wiring a separate MCP server. The skill emphasizes choosing Tavily when built-in tooling is missing or when Tavily’s structured output beats raw HTML fetches. It spans multiple journey phases but lands on idea research as the canonical catalog shelf because that is when most developers first need programmatic web evidence.0installs41Todoist Apitodoist-api is an agent skill that teaches Claude Code, Cursor, and similar agents how to interact with Todoist through the td command-line tool. It documents create, read, update, and delete patterns for tasks, projects, sections, labels, and comments, and it hard-requires confirmation before destructive actions so developers do not lose work from an over-eager agent. Install it when you want implementation plans and bug fixes reflected in Todoist without manual copy-paste. The skill fits developers who already use Todoist as their system of record and want procedural knowledge in SKILL.md instead of ad-hoc API guessing. It is phase-specific to build and PM workflows rather than an methodology.0installs42UltrathinkUltrathink is a minimal agent skill plugin that displays colorful ANSI art spelling “ultrathink” when you say the word or run the /ultrathink command. developers install it as a light-touch ritual or Easter egg inside Claude Code–style sessions: a visual pause that signals deliberate thinking before diving back into implementation. It does not analyze code, change files, or advance the product journey on its own. The skill is intentionally narrow—terminal output only—so it stays out of the way of serious automation unless you explicitly invoke it. Use it when you want a memorable cue in agent-tooling workflows or to reinforce a “slow down and think” habit without opening another productivity app. Skip it when you need substantive planning, security review, or documentation deliverables.0installs43Upstash Redis Kvupstash-redis-kv is an agent skill that lets developers and coding agents talk to Upstash’s serverless Redis-compatible store over HTTP. Install it when you need to save session context, rate-limit counters, feature flags, or short-lived lists during a build or prototype—especially in serverless or edge deployments where running Redis locally is awkward. The skill maps common Redis data types to REST operations so you do not paste ad hoc API snippets from docs each time. It fits agents building SaaS backends, CLI tools with cloud state, or multi-step workflows that must remember intermediate results. Scope is intentionally narrow: key-value and Redis primitives via Upstash, not self-hosted cluster administration or full ORM design.0installs44Youtube Transcriptyoutube-transcript gives agents a clear procedure for pulling spoken content from YouTube videos when the user supplies a valid URL and asks for transcripts, subtitles, or captions. Developers use it to mine conference talks, competitor demos, and tutorial chains without manually copying auto-captions. The skill defines when to activate—YouTube watch or short links—and how to format results, including optional timestamps for citing moments or stripping them for clean summaries and blog drafts. It is a lightweight integration-style skill rather than a full media pipeline: one video, one text artifact. That makes it valuable across research (idea), validation prototypes informed by expert videos, content repurposing, and learning while implementing features shown in screencasts. Shelf placement under Idea → research reflects the most common entry point, not an exclusive limit.0installs