This week we added a new ask-assistant to every page on skillsandagents.co, published a public index of the connectors our catalog supports, and shipped self-serve invite recovery and approval checkpoints for agent runs.
Site Updates
A “Scoped assistant” now sits on every page. It’s an ask box that answers questions about the catalog and the platform, grounded in the site’s own content, with real links back into it. Ask it something off-topic and it says so instead of guessing.
We launched a new public page: /connectors. It lists all 25 integrations, like Gmail, Slack, and Google Drive, that skills and agents in the catalog can use, along with a count of how many published skills and agents use each one.
Discovery call prep is now listed in the catalog. The skill merged into the library back on August 23rd, but it didn’t have a page pointing to it until this week, so it was effectively invisible. It’s live now on /skills/ and shows up in search.
The /for/searcher/ page gets a one-click “install all 17 skills” button, matching the one already on the CFO page.
We also cleaned up a few smaller things: a “Log in” link now appears in the top nav, a stray preview-card graphic came off the CFO page, and a handful of headline and copy tweaks landed on the VC and home pages.
Platform Updates
Self-serve invite recovery is live. If you never got, or lost, your org invite email, a “Didn’t get your invite?” link on the login page lets you request a fresh one, without revealing to anyone probing the form whether an invite exists for a given address.
Agent runs can now stop and wait for a person. When a run hits a checkpoint step, it halts and shows a banner with Approve or Reject buttons for whoever’s on review. Reject ends the run cleanly; Approve resumes it. Every decision is recorded, though there’s no way to browse that history yet. This is step four of our work bundling skills into an agent.
The in-app skills catalog now matches the redesigned public catalog page we shipped last week: search, category, sort, and a list/grid toggle, all combining and surviving an add or remove instead of resetting your filters.
“Add to the brain” moved from a tabbed page to a step-by-step dialog on desktop, walking you through picking a method, filling it in, and reviewing before you commit, without leaving the page.
More settings landed on agent configs: a cadence (weekly, biweekly, monthly), a starting send-approval level, and a recipients list checked against your org’s approved domains. None of it does anything on its own yet, it’s the groundwork for scheduled, gated agent sending.
One thing that isn’t live yet: we built the ability for a running skill or agent to look up your org’s Company Brain mid-run, so it can use real context instead of working blind. It’s behind a flag that’s still off everywhere, so it’s not something you can use today, but it’s built and on its way.
If you are new, start by learning what a skill is or browse the catalog. This update is from the Skills and Agents team. For a behind-the-scenes look at how the week’s work happened, check out Edwin’s weekly log.