Skip to main content
An actor list is a catalog resource that names a reusable set of usernames, grouped by identity provider. A steering policy references it to decide who may steer agents through externally-sourced events such as pull request comments, reviews, and issue or PR triggers:
  • named under the policy’s allowlists, the list admits its actors regardless of the policy’s association tier — this is how you let an approved bot or a trusted outside contributor drive an agent without granting them an organization role;
  • named under the policy’s denylists, the same list refuses its actors, overriding the tier and every allowlist.
The list itself carries no polarity: it is just a named set of actors, and the referencing field decides what it means. The same list may therefore be admitted by one policy and refused by another.
The resource kind is spelled actor-allowlist for historical reasons — it predates the denylists field, and the kind name is part of how stored objects are addressed, so it is kept stable. Read it as “actor list”: the dashboard calls the section Actor Lists, and nothing about a list restricts it to allowing.
Usernames are grouped by provider so the right namespace is matched against the event’s author — a GitHub login is matched in the GitHub OAuth provider, not against an organization or service identity. Matching is case-insensitive. In the dashboard, actor lists are managed under Organization Settings → Actor Lists.

Fields

Entry fields

Each entry in entries groups usernames under one identity provider.

How it is used

An actor list does nothing on its own — a steering policy references it. When a gated event arrives (a PR comment, review, issue-opened, PR-opened, or PR-labeled trigger), the policy decides in this order:
  1. Denylists first. If the author appears in the union of usernames across the policy’s denylists, the event is refused — whatever the tier says and whatever an allowlist says.
  2. Then the tier. The author’s repository association meets the policy’s tier.
  3. Then the allowlists. The author appears in the union of usernames across the policy’s allowlists, matched within the event’s provider.
So an allowlist is what makes a bot account or a trusted outside collaborator able to steer under a tier they would otherwise fail, and a denylist is what stops an actor a broad tier would otherwise admit. A single list may be named by only one of the two fields on the same policy.

Examples

A list of trusted actors, admitted by a policy

A list of unwanted actors, refused by a policy

The same resource shape; only the referencing field differs.
Reference it from a policy’s denylists, and its actors may never steer.

Listing actor lists

Reading a single actor list

Errors

  • steering-policy — references actor lists to admit (allowlists) or refuse (denylists) named authors
  • service-profile — the identity that automated agents commit under
  • Events — the webhook events steering gates
  • murmur set — CLI command for creating and updating catalog resources
  • murmur get — CLI command for reading catalog resources