Skip to content

Add Claude and OpenAI WorkIQ plugin submission packages - #248

Merged
Darrel (darrelmiller) merged 5 commits into
microsoft:mainfrom
slabereemsft:seanla/claude-openai-plugin-packaging
Nov 13, 2026
Merged

Darrel (darrelmiller) merged 5 commits into
microsoft:mainfrom
slabereemsft:seanla/claude-openai-plugin-packaging

Conversation

@slabereemsft

@slabereemsft Sean Laberee (slabereemsft) commented Nov 11, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

  • Add Claude listing metadata and a transparent icon.
  • Add OpenAI portable submission manifests, OAuth client/scope configuration, listing assets, and authored review scenarios. Omit the review recording field under the submission owner's reported video exemption; retain waiver evidence.
  • Reuse the existing WorkIQ 3.0.0 skills unchanged and preserve the Copilot MCP configuration.
  • Keep generated submission ZIPs under ignored artifacts; also ignore root workiq-*.zip outputs.
  • Submission READMEs, PowerShell packagers, and their PowerShell archive regression tests have been removed from this PR. Existing unrelated repository scripts are unchanged. Shared documentation references have been updated.

Review follow-up

  • R1 addressed: The document-summary scenario uses retrieve (grounding) with optional fetch metadata verification; the exact-subject email scenario uses fetch. Missing retrieval or ownership evidence must be reported without implicit ask delegation or invented ownership. The routing corrections remain; the packaged-manifest assertions were removed with the PowerShell tests.

  • R2 / OAuth follow-up: Changed the OpenAI default scope from WorkIQAgent.Ask to WorkIQAgent.Act only, following the review discussion; no .Read scope was added. The provided client ID is unchanged (identified as OpenAI-owned in the review thread). Intended-host import/schema and live OAuth verification remain separate validation gates.

  • R3 remains open pending live testing with Claude: Verify OAuth sign-in, client compatibility, tool discovery, structured reads, grounded retrieval/source links, and confirmation-gated writes on the intended Claude surfaces. This is a validation gap, not an established implementation defect. No speculative Claude OAuth or shared-skill changes were made.

  • Listing keywords: Added meetings, business apps, and bizapps as requested.

Current validation

  • node --test .\tests\workiq-guidance\static.test.mjs .\plugins\workiq\tests\guidance.test.mjs - 286 passed with the latest manifest and guidance changes.
  • Manifest JSON parsed successfully; requested keywords, Act-only scope, and unchanged client ID verified.
  • git diff --check - passed.
  • No remaining source references to the removed packaging/test scripts or submission README paths.
  • No archive packaging regression tests remain in this PR. Earlier packaging-test results in the discussion apply to the now-removed tooling.

Remaining verification

Live OAuth interoperability, tool discovery, and the authored review scenarios have not been established by these offline checks. Marketplace safety scans, policy/legal approval, and actual host validation remain separate. This PR does not publish either plugin to a marketplace.

Add platform-specific listing metadata, assets, ZIP packagers, and offline regression checks while preserving existing Copilot configuration. Omit the OpenAI review recording field under the submission video exemption.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 0bdba339-7586-40db-9039-515e888bf2c3
@slabereemsft
Sean Laberee (slabereemsft) marked this pull request as ready for review November 11, 2026 15:52
Align document and exact-subject email review cases with shared routing, add packaged-manifest regressions, qualify the reported video waiver, and keep submission ZIPs out of source control.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 85562efa-a5e2-44b7-a3c8-105513af93e4
@slabereemsft

Copy link
Copy Markdown
Collaborator Author

Addressed the actionable review feedback in 00522f5:

  • R1 — routing mismatch: Changed the ordinary document-summary case from ask to retrieve (grounding), with fetch only if metadata verification is needed. Changed the exact-subject email case from fetch or ask to fetch. Kept both prompts and the shared skill unchanged. The document case now requires evidence for ownership and reports unavailable retrieval/missing ownership evidence instead of silently delegating or inventing an owner.
  • Regression coverage: Added focused assertions against the manifest inside the generated OpenAI ZIP. The new test reproduced the original document-routing failure before the fix; the corrected packaging suite passes.
  • Video waiver: Qualified the walkthrough as a submission-owner-reported exemption and explicitly required retaining the waiver evidence.
  • Build outputs: Updated Claude build examples to write into ignored artifacts and added a root workiq-*.zip ignore rule. The local Claude ZIP was moved into artifacts; no generated ZIPs are tracked or included in this PR.

R3 — waiting on Claude testing, not claiming resolved. The Claude package intentionally contains the hosted endpoint without copying Copilot-specific OAuth settings. We still need to verify sign-in/client compatibility, tool discovery, structured reads, grounded retrieval/source links, and a synthetic confirmation-gated write in each intended Claude host. Passing ZIP/manifest checks does not prove runtime interoperability. No speculative OAuth or shared-skill changes were made; a Claude-compatible registration/service configuration may be needed if live testing exposes that requirement. The WorkIQAgent.Ask scope suffix alone does not demonstrate that other operations are unavailable.

R2: Left the documented OpenAI OAuth extension unchanged. Intended-host import verification or vendor clarification remains a separate gate.

Validation: Claude and OpenAI packaging suites passed; shared static guidance plus plugin-local tests passed (286 tests); git diff --check passed. Live Claude testing has not been performed as part of this update.

Remove the Claude and OpenAI ZIP packagers and their PowerShell regression tests. Replace stale script references with archive layouts and manual preparation checks while preserving submission assets, routing fixes, and host-validation caveats.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 85562efa-a5e2-44b7-a3c8-105513af93e4
@slabereemsft

Copy link
Copy Markdown
Collaborator Author

Tolga Kilicli (@tolgaki) Updated in 8c7f678: removed all four PowerShell files introduced by this PR (the Claude/OpenAI ZIP packagers and their two archive regression-test scripts). Existing unrelated repository PowerShell scripts are untouched.

Updated AGENTS.md and both submission READMEs to remove script references and automated-check claims. The docs now provide the archive layouts, Claude-specific MCP JSON, and manual preparation/inspection guidance. No replacement packaging implementation was added.

Submission manifests, listing assets, unchanged shared skills, and the R1 routing corrections remain. The packaged-manifest regression assertions are no longer part of the PR because their PowerShell test file was removed. R3 remains open pending live Claude authentication and tool-behavior testing; R2's OpenAI import/schema validation remains separate.

Validation after removal: all 286 remaining static/plugin-local guidance checks passed, git diff --check passed, and no references to the removed scripts remain. The PR description now reflects the reduced scope and current validation.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 85562efa-a5e2-44b7-a3c8-105513af93e4
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot-Session: 85562efa-a5e2-44b7-a3c8-105513af93e4
@slabereemsft

Copy link
Copy Markdown
Collaborator Author

PR is updated.

@darrelmiller
Darrel (darrelmiller) merged commit df79ddf into microsoft:main Nov 13, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants