Vellum Assistant is a personal AI assistant that remembers information about users, learns their preferences, and takes actions across connected apps. It is intended for people who want an assistant that can manage conversations, unfinished work, and proactive notifications over time. The catalogue skills, hooks, instruction, and setting configure or extend how the assistant works.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add vellum-ai/vellum-assistant --skill restaurant-reservationgit clone --depth 1 https://github.com/vellum-ai/vellum-assistantWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/vellum-ai/vellum-assistant/restaurant-reservation)<a href="https://agentmods.dev/skills/vellum-ai/vellum-assistant/restaurant-reservation"><img src="https://agentmods.dev/badge/skills/vellum-ai/vellum-assistant/restaurant-reservation/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/vellum-ai/vellum-assistant/restaurant-reservation"><img src="https://agentmods.dev/badge/skills/vellum-ai/vellum-assistant/restaurant-reservation.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Prompt Injection · line 171 Subtle instructions detected that may alter agent decision-making or introduce hidden biases.Fix: Review content for implicit steering or bias. Ensure instructions are explicit and align with the skill's stated purpose.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00015 | $0.02583 |
| Opus 5 | $0.00008 | $0.01291 |
| Sonnet 5 | $0.00003 | $0.00517 |
| Haiku 4.5 | $0.00002 | $0.00258 |
Grade B, and why
restaurant-reservation scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Subtle steeringmediumPrompt injection
Instructions that bias recommendations or shape behaviour without the user noticing.
- **NEVER tell the user to sign in themselves.** You handle ALL authentication using `assistant browser fill-credential` and `ui_show` for verification codes. How it starts
The opening of the file, as written. The whole thing — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Book restaurant reservations on OpenTable or Resy using browser automation. All browser operations are executed through the assistant browser CLI, invoked via host_bash.
Anti-Loop Guardrails
Each step has a retry budget of 3 attempts. An attempt is one try at the step's primary action (e.g., clicking a button, filling a form, navigating a page). If a step fails after 3 attempts:
- Stop trying. Do not continue retrying the same approach.
- Fall back to manual. Tell the user what you were trying to do and ask them to complete that step manually in the browser. Give them the direct URL and clear instructions.
- Resume automation at the next step once the user confirms the manual step is done.
If two or more steps require manual fallback, inform the user the automated flow is unreliable and offer to provide the remaining steps as text instructions with links.
Booking Flow - Follow These Steps IN ORDER
Step 1: Collect Reservation Details
Before doing anything, gather the following from the user:
- Party size (required)
- Date (required)
- Time or time window (required)
- Location / neighborhood / city (required)
- Restaurant name (optional - if not provided, will search)
- Any preferences (outdoor seating, dietary needs, etc.)
Do not proceed until all required details have been provided.
Step 2: Choose Provider
- If the user hasn't specified a provider, ask them to choose between OpenTable and Resy.
- Do NOT attempt to browse provider sites to check restaurant availability before signing in - authentication is required first. If the user names a specific restaurant but isn't sure which provider has it, pick the more common one (OpenTable) and fall back to the other if it's not found after searching in Step 4.
Step 3: Navigate and Sign In FIRST
This is the most important step. Reservation sites require authentication before booking.
- Navigate directly to the sign-in page.
- For OpenTable, navigate to:
https://www.opentable.com/sign-in - For Resy, navigate to:
https://resy.com/login
- For OpenTable, navigate to:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 190 lines · 15 tokens per session scan B e727182864eb
restaurant-reservation is a skill published in the GitHub repository vellum-ai/vellum-assistant (1,225 stars, last pushed yesterday), licensed MIT. It adds 15 tokens to every session and 2,583 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (subtle steering). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
ha-browser
Hope Agent browser automation — the standard status → tabs → snapshot → act loop, stale-ref recovery rules, and what to do when login / 2FA / captcha / camera-prompt / dialog blocks progress. Load this skill whenever you reach for the browser tool. Trigger on: user asks the agent to open / control / click / scrape /…
macos-use
GUI control for macOS apps via mediar-ai's mcp-server-macos-use. Click, type, scroll, key-press, open apps — driven by accessibility tree, works in non-interactive Claude Code mode. Use this for any Sutando task that needs to drive another macOS application (Safari, Zoom, Mail, Finder, etc.).
lictor-security-check
Pre-release security audit for ANY project — AI-built or hand-written, web or mobile. Scans the codebase for the full range of real-world risks that get apps breached: leaked API & AI-provider keys, exposed configs/secrets, broken auth & access control (IDOR), injection (SQL/XSS/command), SSRF, open databases & cloud…
lictor-rotate
Walks the user through rotating a leaked API key — step by step, provider-specific. Knows the exact URL to visit, the exact button to click, and how to verify the rotation worked. Supports Stripe, OpenAI, Anthropic, Google Cloud / AI Studio, GitHub, AWS, Slack, Supabase, Firebase, Postmark, and generic OAuth providers.
agent-browser
Browser automation using Vercel's agent-browser CLI. Use when you need to interact with web pages, fill forms, take screenshots, or scrape data. Alternative to Playwright MCP - uses Bash commands with ref-based element selection. Triggers on "browse website", "fill form", "click button", "take screenshot", "scrape…
anthropic-webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.