vellum-ai

Use when the user wants to build, scaffold, ship, or edit a Vellum plugin that bundles multiple surfaces (hooks, tools, skills, and more) into one installable package.

Perform a one-time migration from memory v1, to memory v2, which was introduced in 0.8.0.

Migrate from ChatGPT, Claude, OpenClaw, Hermes, Manus, and other AI assistants into Vellum by inspecting their data exports, conversation archives, files, prompts, custom instructions, memory, saved memories, tools, GPTs, workflows, integrations, and…

Ingest a large dataset into memory as a skimmed map. Cold-store the raw files under a workspace imports directory, census them into a slice plan, skim each slice into compact map pages that point back at the raw files, ingest the map with the memory ingest…

Spawn and manage autonomous background agents for parallel work

Create images from a text description, or edit photos and graphics the user provides (remove backgrounds or watermarks, retouch, restyle, in-paint). Can produce multiple variants when the user wants options to choose from.

Create, edit, and delete custom managed skills in the user's workspace. Use whenever the user wants to author a new skill from a description, scaffold a SKILL.md, or remove a skill they no longer need.

Manage assistant voice and system settings

Run a structured competitor teardown — positioning, pricing, product, GTM, and recent moves — into a head-to-head comparison with recommended responses. Use for "analyze competitor X", "how do we compare to", "competitive teardown", or "battlecard".

Build a demand-generation plan worked backward from a pipeline or revenue target through funnel math to a channel and budget allocation. Use for "build our demand plan", "how do we hit our pipeline target", "plan Q_ marketing", or budget allocation.

Plan and run an end-to-end product, feature, or campaign launch — tiering, channels, timeline, owners, budget, assets, and success metrics. Use for "plan our launch", "we're shipping X", "go-to-market for", or campaign planning.

Act as the user's marketing expert for any marketing question or strategy work — marketing help, marketing strategy, growth, getting customers, positioning, demand gen, launches, content, brand, competitive, or marketing analytics — including for a startup or…

Run a structured positioning and messaging exercise (April Dunford method) to produce a positioning statement and messaging hierarchy. Use for "position our product", "fix our messaging", "what's our positioning", or competitive repositioning.

Create and maintain one living marketing/GTM plan document — positioning, channels, content calendar, paid-ads framework, and action items as checkboxes — and keep it up to date across sessions instead of one-shotting strategy into chat. Use for "make a…

Run a traditional search-engine SEO audit — keyword research and intent mapping, on-page analysis, technical checks, content-gap and competitor comparison — ending with prioritized fixes. Use for "SEO audit", "why aren't we ranking", "keyword research", or…

Design visual and graphic assets — logos, social graphics, ad creative, and other media — with on-brand concept directions and export-ready output. Use for "make a logo", "design a graphic / social image", "create ad creative", or "I need brand / media…

Track writing sessions, maintain a daily word count streak, and deliver guilt messages when the user skips. Logs word counts and durations, shows stats, and celebrates milestones. Use when the user wants to start a writing session, set a word count goal,…

Generate a daily writing prompt calibrated to the user's genre and streak, with difficulty that escalates as they build a habit. Saves drafts to a gallery that grows over time. Use when the user asks for a writing prompt, daily prompt, wants to write…

Deliver constructive-but-savage feedback on a piece of prose. Runs programmatic text analysis first (adverb density, passive voice, weak verbs, clichés, sentence rhythm, word repetition, reading level), then roasts the draft using the real numbers as…

Critique a query letter against publishing industry standards. Checks word count, hook, comp titles, bio, agent personalization, and paragraph structure, then delivers professional feedback on hook strength, story clarity, comps relevance, and bio…

Shared persistence layer for the Admin Copilot — provides the admin_copilot_prefs tool that reads/writes the plugin's preferences and tracked-competitor state. Loaded automatically as an include of the admin-copilot-setup, admin-digest, and competitor-brief…

One-time guided setup for the Admin Copilot — a proactive chief-of-staff that runs a morning digest, keeps the inbox triaged (propose-then-confirm), preps the calendar, and delivers a weekly competitor brief. Use when the user wants to set up, enable,…

The Admin Copilot's morning briefing — a chief-of-staff digest unifying today's calendar and meeting prep, an inbox triage summary with propose-then-confirm actions, new competitor deltas, and stale follow-ups. Use when producing the daily/morning admin…

Recurring competitor and market monitoring. Tracks a saved list of competitors, checks their sites and recent news each run, diffs against the last run, and reports only what materially changed — pricing, launches, funding, hiring, positioning. Use for "track…

First-run company setup for the co-founder plugin. Walks the founder through describing their company, team, stage, and top challenges, then stores the profile. Activates when the user first installs the co-founder plugin, says they're starting a company,…

Company-memory tools for the co-founder plugin: company profile, decision log, OKRs, blind-spot analysis, and briefings. Load when the user wants to log, review, or search a strategic decision, set or update OKRs, update company profile details, or asks "what…

Active co-founder mode for thinking through decisions, plans, and strategy. Pushes back, surfaces blind spots, references prior decisions, and challenges the founder's thinking. Activates when the user wants to think through a decision, asks "what do you…

Generate a board-meeting-style weekly briefing summarizing recent decisions, stuck items, OKR drift, and recommended focus for the next period. Activates when the user asks for a weekly review, a status update, "what happened this week," board prep, or when a…

First-run onboarding for the Polyglot language learning plugin. Use when the user first interacts with Polyglot, when they want to set up a new language, or when they say things like "I want to learn Spanish" or "set up language learning." Gathers language…

Instructions for weaving target language into normal conversations when the user is in ambient mode. Use when the Polyglot plugin is in ambient mode and the user is having a normal (non-practice) conversation. This is the differentiated experience — language…

Instructions for immersive mode where the assistant responds primarily in the target language. Use when the user is in immersive mode — advanced learners or intensive practice periods. The assistant uses English only for clarification or when the user is…

Run a dedicated language practice session. Use when the user asks to practice, says "let's practice [language]", wants to do a lesson, or when a practice session should start. Generates life-contextual scenarios, adapts difficulty, and teaches through…

Coach the user on building sustainable health habits around sleep, daily movement, consistency, and behavior change. Use when the user mentions struggling with consistency, motivation, sleep, getting enough steps, or wants to build healthier routines. Also…

Guide the user on nutrition, meal planning, and dietary choices. Use when the user asks about what to eat, protein intake, calorie targets, meal timing, hydration, macronutrients, or building a meal plan. Also activates when the user mentions struggling with…

Help the user plan workouts, choose exercises, and structure training routines. Use when the user asks about workout splits, exercise selection, progressive overload, rest days, or wants a training plan. Also activates when the user mentions wanting to start…

Reconstruct the history behind a decision, project, claim, or recurring idea using Threadkeeper's staged evidence retrieval. Use when the user asks what happened, why a decision was made, when an idea started, what changed, or where a claim came from.

Review Threadkeeper open loops, memory-gardener proposals, contradictions, and serendipity connection hypotheses. Use when the user asks what needs review, wants to accept, reject, snooze, complete, archive, explore, or dismiss a Threadkeeper item.

Configure Threadkeeper, including an explicit, changeable conversation backfill scope, serendipity setting, and sensitive-category privacy default. Use when the user asks to set up, enable, configure, or backfill Threadkeeper.

Verify native iOS behavior of the vellum-assistant Capacitor app on a real iOS simulator: build from a worktree, boot against the qa-mock-backend server, drive with XCUITest, and record video proof. Use for anything WKWebView/Swift-side — edit menus, native…

Visually test vellum-assistant clients/web changes on a feature branch: boot the app locally with mocked platform APIs, drive it in headless Chromium with mobile/coarse-pointer emulation, dispatch real CDP touch gestures, and produce screenshots plus an…

Stand up a fake vellum platform + daemon backend so any vellum-assistant client (web via Playwright, iOS via simulator, Electron) boots to the real chat UI with scripted conversation state — no real accounts, daemon, or network. Shared core of the qa-web-ui…

Set up the iMessage channel with the user's own Photon or Comms by Osis account so the assistant can send and receive texts. Use when the user wants to text the assistant, when a send fails with a missing credential or 401, or when the channel reports it is…

Send an iMessage or SMS to a phone number. Use when the user asks the assistant to text someone, or asks it to text them.

Answer board-game rules questions from the installed boardgame-rules plugin, or validate a specific play situation against a pre-authored worked example. Use when the user asks how a supported game works, what a rule is, whether a move is legal, or whether…

Teach a supported board game at the table, during one sitting: ask what the players already know, teach in layers (goal, turn loop, exceptions), cite every ruling, and bridge to their known games — but only when the corpus has a real analogy. Use when someone…

Send a note-taking bot into a meeting via Recall.ai; only when the user explicitly asks.

Guide the user through setting up the Recall.ai API key for the Meeting Bot plugin.