Agent skills

The pillar: AI-native, agent-first workflows

Most product tools generate admin work for the human. Tickets to update, dashboards to maintain, statuses to sync. The work becomes keeping the system fed. AI-native, agent-first workflows invert that. The tools do the work. The human steers the strategy.

This is one of the pillars of [The Revelica Method](/method). Agent skills are how it shows up in practice. When you run a skill, the agent does the synthesis. You stay in the loop on the judgment calls.

What a skill looks like

Every skill has three properties that matter:

• It does one focused thing. Not a chatbot, not a generalist. "Map customer experience" reads an interview transcript and produces an experience map with opportunities. That's the whole job.
• It produces structured output. Skills don't hand back conversational text. They produce typed artifacts (Hypothesis, Assumption, Evidence Source, Opportunity, Story Map, Competitive Analysis) that the rest of the system, and your coding agent, can use directly.
• It cites its work. When a skill makes a claim about a customer, it points at the source. No fabrication. When the receipt isn't there, the skill surfaces the gap.

How skills chain into playbooks

A playbook is a chain of skills wired together to handle a bigger job end to end. The agent runs each skill in order, passing outputs along so the next step has the context it needs.

Example: the Set up workspace playbook takes a single company URL and runs nine skills in sequence. Research the company, define the customer segment, articulate the value proposition, find and research competitors, surface opportunities, and so on. What would be a week of manual setup becomes a single playbook run.

Browse the current library

The full set of agent skills and playbooks lives in the template library. The library is the canonical source. New skills land regularly as new patterns prove worth automating.

FAQ

Can I write my own agent skills?

Custom skills are on the roadmap. Today, the agent runs the skills in the template library. If there's a workflow you'd want as a skill, tell us. The library grows based on what customers actually need.

How does this differ from running an LLM directly?

A direct LLM call gives you conversational text. An agent skill gives you a typed, cited artifact that fits into the rest of your discovery work and that your coding agent can consume. The skill also enforces structure and evidence discipline a free-form prompt does not.