RC
An AI agent skill is a packaged, repeatable task an agent can run on demand, such as researching an account, drafting a follow-up or updating a deal, with its instructions, the data it needs and the rules it must follow. A playbook (or play) chains several skills into a recurring GTM motion, like "new inbound lead to booked meeting". Skills make agents consistent. Playbooks make them useful at the level of the motion, not just the task.
What is an AI agent skill?
A skill is how you teach an agent to do one job your way, every time. It usually bundles four things:
Instructions: the steps, the standard, the tone.
Context: the data it should read (the account, the deal, your ICP, past calls).
Tools: the systems it may use, such as the CRM, the call recorder or email.
Rules: what it can do alone, and what needs a person's approval.
The idea has gone mainstream fast. Anthropic introduced Agent Skills in October 2025, describing them as folders of instructions, scripts and resources an agent loads only when the task needs them, and released the format as an open standard in December 2025. Most GTM teams don't write skills as files, but the principle carries over: write the job down once, and every agent that runs it does it the same way.
How is a skill different from a prompt, a workflow or an agent?
Prompt: a one-off instruction, written fresh each time. Example: "Summarise this call."
Skill: a reusable task with instructions, context, tools and rules. Example: "Post-call summary: pull the deal, summarise against MEDDICC fields, propose CRM updates for approval."
Workflow: a fixed sequence of steps that runs the same way every time. Example: "When a form is submitted, create a lead and notify the owner."
Agent: software that decides which skills to use, and when, to reach a goal. Example: a follow-up agent that picks the right skill for each open deal.
Playbook: a chain of skills that runs a whole GTM motion. Example: "Inbound lead to booked meeting."
The short version: an agent is the worker, skills are what it knows how to do, and a playbook is the job. Plenty of products sold as agents are really workflows, which we unpacked in most of your agents are just workflows.
What are examples of AI agent skills for GTM teams?
Ten that show up across sales, marketing and RevOps:
Account research brief. Pulls firmographics, recent news, hiring signals and past touches into a one-page brief before outreach.
ICP fit scoring. Scores an account or lead against your written ICP and explains the score.
Personalised first touch. Drafts an opening email grounded in the research brief and your messaging, for a rep to approve.
Meeting prep. Builds a pre-call brief: attendees, open opportunities, last conversation, likely objections.
Post-call summary and CRM update. Summarises the call against your qualification framework and proposes field updates for approval.
Follow-up draft. Writes the follow-up from the call notes, with the agreed next step.
Stalled deal check. Flags opportunities with no activity, a missing champion or a slipped close date, and suggests the next move.
Forecast note. Drafts the weekly deal-by-deal commentary from CRM and call data.
Data hygiene fix. Finds duplicates, missing fields and contacts who have changed jobs, and proposes the fixes.
Win/loss digest. Summarises why recent deals were won or lost, from call transcripts and notes.
Notice that half of them write to a system of record. Those are the skills that need approval rules from day one. See what governance AI agents need on a revenue team.
What are examples of AI agent playbooks?
A playbook strings skills together around one outcome:
Inbound to meeting: ICP fit scoring → account research brief → personalised first touch → meeting booked.
First meeting to next stage: meeting prep → post-call summary and CRM update → follow-up draft.
Deal rescue: stalled deal check → account research brief (what changed?) → follow-up draft to a new contact.
Weekly pipeline review: data hygiene fix → stalled deal check → forecast note.
Closed-lost to learning: win/loss digest → updates to the ICP and messaging that every other skill reads.
The last one matters most. A playbook that feeds what it learned back into the shared context is how agents get better over time, not just faster. Teams are moving this way: KPMG's Q2 2026 AI Pulse found the share of organisations orchestrating multiple agents across workflows doubled from 9% to 18% in one quarter.
How do you build good GTM skills?
Start from a task people already do well. Write down how your best rep or RevOps lead does it. That becomes the skill's instructions.
Name the outcome. Each skill should exist to move something: a meeting, a stage, a cleaner record. If it doesn't, cut it.
Ground it in shared context. A skill is only as good as the data it reads. If every agent keeps its own copy of the ICP, your skills will disagree. We covered why in why AI sales agents don't learn from each other.
Set the write rules. Read-only by default. Approval for anything that changes the CRM or contacts a buyer.
Write it once, use it everywhere. The same "post-call summary" skill should run whether the call was handled by your agent, a vendor's agent or a rep using an assistant.
Learn from the edits. When people change a skill's output before approving it, that change is the next version of the instructions.
Chain only proven skills. Get two skills reliable on their own before you string them into a playbook.
What should you look for in tooling?
Can skills be shared across agents from different vendors, or only inside one product?
Do skills read from one shared, structured source of GTM truth?
Can you set per-skill permissions and approval rules?
Are edits and overrides captured to improve the skill?
Can skills be chained into plays, with a record of each step?
Does it connect through an open standard such as MCP, so each system is wired once?
For the full vendor checklist, see our guide to evaluating GTM agent platforms.
Where does wysdym fit?
wysdym is the operating layer for agentic GTM. It grounds every agent in your GTM truth, runs the motion under your governance, and gets sharper with every deal you close. Skills is one of its five pillars: a shared library of GTM skills that agents invoke over MCP through the Gateway, so any agent, from any vendor, runs the same task the same way. Every skill reads from Cortex, the shared GTM knowledge graph, and runs under Governance, with per-agent permissions, approval queues and an audit trail. Plays, chained skills that run recurring motions, are coming. More on the category in what an agentic GTM platform is.
If you want to see what this looks like on your own stack, book a conversation or see how it works at wysdym.ai.
FAQ
Are AI agent skills the same as plugins or integrations?
No. An integration gives an agent access to a system. A skill tells it what to do with that access, how well, and within which limits. A skill usually uses one or more integrations.
Who should write GTM skills: RevOps, enablement or engineering?
The people who know the job write the instructions: usually RevOps and enablement, with the best reps as the source. Engineering, or your platform, handles the connections and permissions.
How many skills does a GTM team need?
Fewer than you think. Start with the three or four tasks that happen every day and touch revenue, such as research, meeting prep, post-call updates and follow-up. Add more only once those run reliably.
What's the difference between a playbook and a sales sequence?
A sequence sends a fixed series of messages on a schedule. A playbook chains skills that read the situation and decide the next step, and it can include research, CRM updates and handoffs, not just emails.
