In Flow Core, an Instruction can be understood as a single executable action inside a node. It is not a node itself, but the smallest runtime unit a node uses to perform work.

A Node defines graph-level behavior such as scheduling, input and output flow, Guard handling, busy-state behavior, and BT / FSM / Stack semantics. An Instruction, by contrast, defines what actually happens after execution enters that node.

Mechanism

• In the editor, what gets stored is a FlowInstructionEntry, which consists of id + SerializeReference data + enabled + comment. It does not serialize the runtime instruction object directly.
• At runtime, the instruction interface is IInstruction.ExecuteAsync(FlowContext context), which returns ValueTask<InstructionResult>. This means an instruction can either execute synchronously or wait asynchronously.
• InstructionData is the editor-side data container. It is responsible for methods such as Clone() and Summary(). The actual execution logic lives in the class that implements IInstruction.
• Instructions are not resolved at runtime through reflection. Instead, they rely on a registry + code generation pipeline. InstructionRegistry stores the mapping from id -> factory, GeneratedInstructionRegistry collects the generated entries, and FlowRegistryBootstrap performs the unified registration at startup.
• When a node is compiled, FlowRuntimeInstructionBuilder converts the entry list into an IInstruction[]. Disabled entries are skipped, and unregistered ids will produce warnings.
• At execution time, Action / Loop / BT / FSM / Stack each run their corresponding IInstruction[] in sequence. InstructionResult then feeds back into node-level control flow, so it does more than simply indicate “success” or “failure.”

The meaning of Instruction Result is especially important:

• Continue Continues executing the remaining instructions, or allows the node to finish normally.
• Break Immediately stops the current instruction list, and the node exits through Out.
• Restart Restarts execution from the beginning of the current instruction list.
• LoopContinue In a normal Action node, this is generally equivalent to continuing execution. In Loop style nodes, it means skip to the next iteration.

Usages

Commonly used instructions are typically general-purpose gameplay or application actions. In most cases, they return Continue and do not directly manipulate Flow Core’s scheduling system.

• Set* / Set List* Write values or lists into the Blackboard. This is one of the most commonly used instruction categories.
• Log* / List LogCount* Used for debug output and inspection.
• Math* / Bitwise* / Make* Handle numeric calculations, vector construction, and struct assembly.
• List Operations* / List Rollover / Shuffle* Perform list manipulation such as add, remove, update, query, sort, reverse, and shuffle.