When an issue arrives, your AI Support Agent reads the context and matches it against your Skills library. If a Skill applies — say, "bug-report-triage" for an issue that includes error messages and unexpected behavior — the agent invokes it automatically. No manual configuration per issue. No training required.
Skills are injected into the agent's context at the start of every issue. The agent reads the name and description of each Skill, decides which ones apply, and loads the full instructions only when needed. This keeps the agent focused and avoids overwhelming it with irrelevant steps.
In the Assist Agent panel, human agents can type / to search and trigger any Skill from the org's library. The Skill loads its instructions inline and runs through the process step by step — the same way a great senior rep would walk through a ticket.
Create a Skill three ways: solve an issue in Assist Agent and tell it to turn what you did into a repeatable Skill. Paste in an existing runbook or SOP and ask Pylon AI to convert it. Or import a Skill you've already built in another tool. No writing from scratch. No engineering required.
Every Skill you create is available to your entire org — to all AI agents and all human agents. Skills are versioned and can be organized in folders, so you can iterate safely and clearly define permissions and structure.
Skills aren't just for AI. Human agents in the Assist Agent panel can search and trigger any Skill with a / command. The same process that your Support Agent runs autonomously at 2am is the same process your human agent runs at 10am. Consistency across your whole team, not just your AI.
Skills can call Connectors — any MCP-compatible tool or APIs your team uses. A Skill can call your internal API, create a Linear ticket, check a Stripe subscription, query your codebase, or read a Notion doc. You define the process; Connectors give it reach.
.png)
Skills are named, versioned sets of instructions that tell an agent, AI or human, how to handle a specific type of situation. Each Skill has a name, a description of when to use it, and step-by-step instructions written in plain English. Think of a Skill as a modernized runbook: more flexible, invokable by both AI agents and human agents, and built on an open standard so your processes are not locked into any single tool.
AI agents invoke Skills automatically. When an issue arrives, the agent reads its context, matches it against your Skills library, and runs the right Skill without any manual routing. Human agents can search and trigger any Skill with a / command inside the Assist Agent panel. The Skill loads its instructions inline and walks the agent through the process step by step. The same Skill your Support Agent runs autonomously at 2am is the same one your human agent runs at 10am.
There are three ways to create a Skill, all without writing one from scratch. You can solve an issue in Assist Agent and tell it to turn what you just did into a repeatable Skill. You can paste in an existing runbook or SOP and ask Pylon AI to convert it. Or you can import a Skill you have already built in another tool. No engineering required, no blank page.
Every Skill you create is available to your entire org, to all AI agents and all human agents. Skills are versioned and can be organized in folders, so your team can iterate safely with clear permissions and structure. One library, shared across every agent.
Yes. Skills are composable and can call Connectors, any MCP-compatible tool or API your team uses. A Skill can call your internal API, create a Linear ticket, check a Stripe subscription, query your codebase, or read a Notion doc. You define the process in plain English, and Connectors give it reach across your stack.
