This week’s AI contributor meeting covered WordPress 7.1 RC1, WordCamp US contributor table and prompt bar planning, a proposed “trust ladder” concept for AI agent permissions, and an open structured content and abilities proposal awaiting review.
WordCamp US: Contributor Table and Prompt Bar Planning
@jason_the_adams, @nikskyverge and @justlevine have a separate call tomorrow to finalize the AI contributor table. @flixos90 is expected to join partway through Contributor Day.
This is the first year the team is running a “prompt bar,” and planning focused on how to make it useful for both attendees and the team:
- Variety of tasks and formats — ensuring there’s something for contributors without laptops (design/writing tasks, not just code), following feedback from previous WordCamp Europe events.
- Collecting community feedback — rather than one method, the team wants multiple ways for people to contribute: talking directly to volunteers, writing on post-its, filling out a Google form, or posting in the Slack channel. A shared spreadsheet will let table volunteers log conversations and ideas in real time so nothing gets lost.
- Per-topic QR codes — instead of a single QR code, the team leaned toward separate codes for abilities, the API, connectors, and the AI Client, each with a short description. This doubles as a lightweight way to see which topics attendees are most curious about.
- Reference materials — @mikeyarce is putting together a two-sided info sheet (what the team is working on, how to get involved, common questions, links) for volunteers and attendees.
- Broader ideas raised — advancing the AI contribution guidelines from V0 to a V1 that applies project-wide (not just within the AI Team), running a focused sprint on the skill-review backlog, and updating WP Bench to compare performance with and without skills.
DECISION: Set up a shared spreadsheet for table volunteers to log community feedback and conversations during Contributor Day and at the Prompt Bar table.
A thread will go into the Slack to collect further ideas asynchronously before the team acts on planning next week.
AI Trust Ladder
@jason_the_adams opened a discussion on a “trust ladder” concept currently being discussed internally at Automattic, and how it might apply to WordPress Core.
The idea: define four levels of trust (plus a zero/kill-switch level) that describe how much autonomy an AI agent has been given, similar in spirit to WordPress user roles (subscriber through administrator). Abilities could optionally declare the trust level they require, and a site owner could set their own trust threshold, which would determine which abilities are exposed to an agent at all.
Key points from the discussion:
- @ovidiu-galatan noted WordPress.com already implements something similar: users can enable or disable individual abilities, granular read/write permissions exist for external agent access, and destructive actions always require explicit approval regardless of trust settings.
- @jorgefilipecosta raised the implementation question of how a trust level would map to what’s registered with MCP; the working answer was that abilities above a site’s trust threshold simply wouldn’t be registered with the agent at all.
- @nikskyverge distinguished between the trust level an ability’s author declares it needs, and a separate, site-owner-controlled “agentic trust level” — the latter being the new, user-facing control this proposal introduces.
- @karmatosed cautioned against presuming a user’s level of trust in either direction, flagged emerging disclosure requirements for AI-generated content in some countries, and argued this should live in Core (kept simple) rather than plugin territory, with an eye toward audit trails and transparency becoming expected.
- The group discussed whether an AI agent should be represented as a distinct entity or as a new type of user — leaning toward “type of user,” partly because that’s the direction regulation seems to be heading.
- @karmatosed suggested shipping with predefined roles as a starting point, reusing WordPress’s existing role language (contributor, editor, administrator) rather than inventing new terminology.
This is an early-stage discussion rather than a proposal moving into implementation, and the team welcomes further thoughts.
Structured Content and Abilities Context Proposal
@jeffpaul flagged an existing proposal on structured content context abilities that hasn’t gotten much visibility yet. No discussion was needed beyond making sure the team has seen it and can react if needed. It will also be raised in tomorrow’s biweekly Slack meeting for broader visibility.
Next Steps
- Set up the shared feedback-logging spreadsheet for the WordCamp US prompt bar table.
- Continue prompt bar planning async in the Slack channel; revisit next week.
- Finalize the AI contributor table plan (@jason_the_adams, @nikskyverge, @justlevine call tomorrow Thursday, Aug 6th).
- Continue discussion on the AI trust ladder concept and how it could map to Core.
- Review and react to the structured content and abilities proposal; raise again in tomorrow’s Slack meeting.
- Embeddings & Streaming support ideally continue to get testing and reviews to ensure those are stable and released, with hope of getting into WP 7.2 later this year.
Upcoming Meetings
- Contributors are welcome to join the weekly AI contributor call. All team meeting schedules are published on the WordPress Meeting Calendar.
- AI Team Office Hours (Slack): Thursday, 6 August 2026, 17:00 UTC. Facilitator: @jorgecosta.
Props to @jeffpaul for pre-publish review.