AI
Agent Skills: How to Give Your AI Your GTM Playbook
An agent skill is a folder with a SKILL.md file that teaches an AI agent how to do a specific task: when to use it, what to ask for, the steps, the output and the rules. It is an open format, originally developed by Anthropic and now supported by Claude, ChatGPT Codex, Cursor, GitHub Copilot, Gemini CLI and many others. For GTM teams, it turns repeatable workflows into playbooks any agent can run.
Key Takeaways
- A skill is written once and loaded whenever a task matches, so a playbook doesn't depend on someone remembering the prompt.
- The format is open: the same SKILL.md works across many AI agents, not just one vendor's.
- Good GTM skills encode judgment as rules: what inputs to require, what never to invent, what the output must contain.
- Start with the workflows your team repeats every week.
Prompts don't scale. Playbooks do.
Most teams use AI the same way: someone writes a good prompt, gets a good result, and the prompt lives in their head or a chat history. The next person starts from scratch, and the output varies with whoever asked.
An agent skill fixes that. It is a small, version-controlled file that describes how a task should be done, and the agent applies it every time the task comes up.
What a skill actually is
At its simplest, a skill is a folder containing a SKILL.md file: a short header with a name and a description of when to use it, followed by plain-language instructions. It can also bundle scripts, templates and reference documents.
Agents load skills through progressive disclosure. At startup they read only each skill's name and description; when a task matches, they read the full instructions. So a team can keep many skills installed without cluttering every conversation. The format is documented at agentskills.io (opens in a new tab).
Where skills work
The format was originally developed by Anthropic and released as an open standard. Its client showcase lists support in Claude, Claude Code, ChatGPT and Codex, Cursor, GitHub Copilot and VS Code, Gemini CLI and many other agents.
That matters for GTM teams, whose people rarely use the same AI tool. A skill written once can follow the workflow, not the vendor.
What makes a good GTM skill
- A precise trigger. The description should say exactly when to use it: 'score and tier a list of target accounts', not 'help with sales'.
- Required inputs. Name the data the agent must ask for, such as closed-won history, a CRM export or a segment hypothesis, and what to do if it's missing.
- Ordered steps. The same sequence a good operator would follow, so results are comparable across runs and people.
- A defined output. One document with named sections, so the next step (a rep, a sequence, a CRM import) can use it directly.
- Rules. Never invent metrics, names or contacts; label assumptions; cite evidence. This is where team judgment becomes repeatable.
How Clarafold publishes its workflows
Every Clarafold workflow is published twice: as a step-by-step guide to read, and as an agent skill to download. ICP research, outbound personalisation, content research, lead enrichment and account prioritisation each come as a SKILL.md with the inputs, steps, outputs and rules above, and the library grows as new systems are built.
Install one in your own agent and ask it to run the workflow on your data. If you'd rather have the whole system designed and operated for you, that's what GTM Strategy and the other services are for.
Put It To Work