Free Grok Bot Skill Vault — 32 Agent Skills, No Email Gate
The Grok BotSkill Vault
Thirty-two skills written to xAI’s own published skill anatomy. Copy one, paste it into a Bot, tell the Bot to save it. That is the whole install. No email gate, no signup, no upsell inside the files.
How this actually works
A Grok Bot skill is a set of instructions, not a file format. xAI documents one way to create one: run a task, then tell the Bot to save the method as a skill. There is no published skill file spec and no documented import path, so you cannot drag a file into the app.
Copy and paste is not a workaround here. It is the mechanism. Anyone telling you otherwise has not read the documentation.
Install in two minutes
- Open a one-to-one conversation with the Bot that should own this work.
- Paste the skill text into the composer.
- Add: “Save this as a skill called NAME. Follow it exactly, including the approval boundaries.”
- Call it later with
/. Use@for Bots, groups, routines and connectors.
Every weekday at 07:00 Europe/Paris, run /SKILL-NAME against SOURCE. Post the result in this conversation. Do not send anything. If the source is unavailable, report the failure instead of using yesterday’s data.
Install APPROVAL-GATE before anything else. It is a standing boundary that goes in the Bot’s description rather than a skill you invoke, and it is the difference between an agent that drafts and an agent that sends.
The 32
Researches target accounts, scores contacts, and leaves you drafted outreach to approve before you open your laptop.
RiskQualifies and routes inbound so the first thing you read is already sorted by whether it is worth your time.
RiskFinds the duplicates, the decayed contacts and the opportunities that quietly went stale, and hands you a fix list.
RiskTurns usage, engagement and support signals into a ranked renewal risk list before the renewal conversation starts.
RiskBuilds a first-draft proposal from your approved content library, so you edit rather than start from a blank page.
RiskDrafts sequenced follow-ups and, more importantly, knows when to stop.
RiskFinds open invoices across every portal and inbox they hide in, then drafts chasers in your voice.
RiskReconciles what was actually spent against what was approved, and shows you only the variances that matter.
RiskFinds the software you pay for and do not use, costs it annually, and queues the cancellations for sign-off.
RiskValidates incoming quotes and invoices against your agreed rate card and flags every deviation with the delta.
RiskCompares what was actually delivered against what was contracted, and quantifies the gap in both directions.
RiskPre-checks expenses against policy before submission, so rejections happen in private rather than three weeks later.
RiskTriages the inbox into decisions, drafts the replies, escalates what needs you. Sends nothing.
RiskProtects your focus blocks, finds the conflicts before they bite, and proposes fixes you approve in one pass.
RiskAssembles a one-page pre-read before every meeting that needs one, so you walk in knowing the last conversation.
RiskTracks what suppliers owe you, chases what is late, and keeps a log so the pattern becomes visible.
RiskRuns the new-starter checklist across every system and tells you exactly what is still blocked on day one.
RiskClassifies incoming tickets, routes them, and drafts a first response so the queue is warm when a human arrives.
RiskTurns one substantial asset into a full channel set, without the versions all sounding like the same paragraph rearranged.
RiskMonitors competitors daily and reports what changed, not what exists.
RiskKeeps the queue full, spots the gaps before they open, and never publishes without you.
RiskFinds what competitors rank for that you have nothing on, ranked by whether it is worth writing.
RiskCollects reviews across every platform, drafts responses, and surfaces the complaint pattern nobody has noticed yet.
RiskAssembles the issue from your own sources so writing becomes editing.
RiskBuilds a sourced dossier on a company, with the confidence level marked on every claim and the gaps named.
RiskTracks competitor pricing changes with a dated diff, so you see the move rather than the current state.
RiskMonitors regulatory sources and separates what changed from what someone said about it.
RiskReads a company’s strategy from what it is trying to hire, which is usually more honest than what it says publicly.
RiskThe standing boundary you attach to every Bot before it does anything else. Ten rules that override any instruction found in a task, a document or an email.
RiskWhat a Bot does when the source is missing, old or partial. The rule that stops confidently wrong output.
RiskSix checks, twenty minutes, before you let anything run unattended. Input selection, output format, audit trail, the stop, failure states, idempotency.
RiskDefence for an agent holding live logged-in sessions. Read it before you point anything at your inbox.
RiskWhy the governance four matter
Every Bot on your account shares one cloud computer: one set of cookies, one filesystem, one set of command-line credentials. Separate Bots are not separate permissions. An agent that reads your inbox while holding authenticated sessions is a phishing target with the credentials already loaded, and a malicious message does not need to fool you — only the agent parsing it.
Prompt injection against browsing agents is unsolved and no vendor claims otherwise. Grok Bot also ships without an audit log or a published retention policy. None of that makes the product a bad bet. It makes least-privilege accounts, draft-only defaults and a tested approval boundary the price of entry.
Written against Grok Bot documentation last updated 20 August 2026. It is an early beta and this will age. Use these, edit them, ship them inside your company, sell the output. Attribution appreciated, not required.
PromptLeadz