FollowAgents AI AGENTS DIRECTORY
Home / AI Agent Glossary / Playbook
Skills & Workflow

Playbook

A playbook is a set of “if this situation, apply this rule, take this action” instructions written for a recurring class of problems, rather than a single fixed sequence.

SituationDecision ruleAction
A playbook maps a situation to a rule and an action.

Some processes really are linear: step 1, step 2, done. Others depend on what's actually happening — a support agent's response to “user can't log in” should differ from its response to “user was overcharged.” A playbook is written for that second kind of situation: it pairs recognizable situations with the decision rule and action to take for each one, the way a human team's runbook or escalation guide works.

The simplified diagram on this page shows a single situation flowing to a single decision and a single action, but that's a simplification for illustration. A real playbook almost always branches — several possible situations, each routed to its own rule and action — which is closer to what a dag-workflow actually models than a straight line.

Playbooks are usually written in prose or as structured conditional rules (agents-md-style instruction files are a common home for them), rather than compiled into a fixed program. That makes them easy for both humans and agents to read and update, at the cost of being less strictly enforced than a hard-coded pipeline.

Example

A deployment playbook might read: “if the test suite fails, block the merge and notify the author; if tests pass but the change touches /auth, require a second reviewer before merging; otherwise, merge automatically.” Each clause is a situation paired with a rule and action.

How it differs

Playbook vs. workflow: a workflow is often written for one specific sequence of steps; a playbook is written to cover a family of situations, with different branches for different cases, closer in spirit to a decision tree than a fixed line of steps.

Common misconceptions

Often assumed: A playbook is just documentation that people read.
Actually: In an agent context a playbook is often meant to be followed programmatically or read directly by an agent, not only by humans.
Often assumed: A playbook, like the diagram here, is a single straight line.
Actually: Most real playbooks branch on the situation — multiple conditions each routing to a different action — a single-path playbook is the exception.

FAQ

What is a playbook for an AI agent?
A set of situation-to-action rules — “if X happens, do Y” — written to cover a recurring class of problems rather than one fixed sequence.
How is a playbook different from a workflow?
A workflow is typically one sequence of steps; a playbook branches, mapping several possible situations to different rules and actions.

Last checked: 2026-08-28

Related terms

Workflow

A workflow is a defined sequence of steps — carried out by an agent, a human, or both — that runs from some starting trigger to a finished result.

DAG Workflow

A DAG workflow is a workflow modeled as a directed acyclic graph — steps that can branch into parallel paths and later merge back together, with no path ever looping back on itself.

AGENTS.md

AGENTS.md is a plain-text file placed in a project's repository that gives AI coding agents project-specific instructions, like how to set up, build, test, and follow the project's conventions.