Griptape Combine: Rules List

Griptape agents behave how their rules tell them to behave - and one ruleset is rarely enough. Griptape Combine: Rules List takes several rulesets and bundles them into a single RULESET you can hand an agent. Think of it as the "personality stack" node: system prompt stuff, formatting rules, tone constraints, refusal boundaries, all bundled up and passed along together.

What it's for

In the Griptape model, a Rule is a single instruction ("Always be kind and friendly") and a Ruleset is a named bundle of them. The Create Rules node makes one ruleset; this node combines several of those into a bigger set. That's how you build an agent that's both helpful and formatted and safe without cramming every instruction into one text box - you make one ruleset per concern and merge them.

You then wire the combined RULESET into the rulesets input on Griptape Create: Agent (or the Create Agent from Config path). The README frames it as "control agent behavior and personality with access to Rules and Rulesets," and this is the node that makes multi-ruleset agents practical.

How it works

Each input is a RULESET (which is itself a list of rulesets from the source's perspective). The node flattens them:

rule_list = []
for rule in rules:
    rule_list.append(rule[0])

So it takes the first ruleset from each connected input and collects them into one list that the agent treats as its full ruleset set. Connect rules_1, and as with all the combine nodes, more inputs appear the more you connect.

Inputs and output

• rules_1 (RULESET) - a ruleset to add. Connect outputs from Create Rules (or cloud rulesets) here.
• Output: RULESET - the combined set, ready for an agent's rulesets input.

Installation

Ships in ComfyUI Griptape Nodes:

• ComfyUI Manager: search "Griptape" → install.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/griptape-ai/ComfyUI-Griptape, restart ComfyUI.

Standard pack install - griptape[all] plus the rest, with the usual torch reinstall note in the README if the environment fights back.

Gotchas

• Rules are appended in connection order, so if two rulesets contradict each other, the later one tends to win in practice. Order your inputs with that in mind.
• This node does no LLM work itself. If your agent ignores the rules, it's not this node's fault - check that the combined ruleset is actually wired into the agent, and that your model isn't just weak at following instructions.

For anyone building a real agent rather than a demo, this is the node that keeps the behavior definition organized. Rules in, personality out.