Daily Edition Sources +4

Benjamin Shafii Put OpenWork's Docs Inside The Agent

A June 15 OpenWork patch turns product documentation into agent tools, making app help a runtime behavior instead of a repo archaeology task.

Diagram Punk poster showing a user question flowing through OpenWork docs search and docs read cards into a cited answer, with a Benjamin Shafii commit card and an impact-not-measured caveat.
Diagram Punkthe docs moved into the agent loop.
repo different-ai/openwork evidence
4 source signals 1 repo commit 3d03303
Evidence: commit 3d03303 / June 17, 2026 / Daily Edition
Open Edition Evidence below

Benjamin Shafii's June 15 commit 3d03303 is a maintainer move with a human target: when an OpenWork user asks about Slack, permissions, providers, or Cloud MCP, the agent should search bundled docs before spelunking code.

The Public Trail

The patch changes apps/server/src/opencode-plugins/openwork-capabilities-knowledge.ts, adds schemas for openwork_docs_search and openwork_docs_read, builds a docs cache, scores pages by query terms, returns excerpts, and reads docs by path.

The test is concrete. It asks for Slack connection guidance, expects the bundled docs path start-here/connect-your-stack/connect-slack-mcp.mdx, and checks for OAuth callback and scope details.

Why This Contribution Matters

OpenWork's product thesis depends on less technical users trusting a complex agent surface. Docs lookup moves support knowledge into the same runtime where the user is asking for help, while telling the agent to cite docs and infer from code only when docs are missing or stale.

The Conversation To Open

The fair question for Shafii and the OpenWork maintainers is how this stays current. Bundled docs help onboarding, but stale docs inside an agent can become confident wrong guidance unless releases, tests, and docs ship together.

Evidence Trail

Receipts below the story

The article above is the public narrative. This section keeps the source trail, limits, and reporting notes on the same page.

Edition
DateJune 17, 2026
LaneDaily Edition
Confidence78%
Sources4
Reposdifferent-ai/openwork

Primary Evidence

Evidence Limits

  • The article does not claim the docs lookup improved measured user outcomes.
  • The evidence shows a contributor-visible implementation and test, not a public usage metric.
Letters & Corrections

Send a note to the desk

Corrections, missing context, or a follow-up lead.