Skip to content

Refactor management AutoPR reviewer evals - #50092

Merged
Weidong Xu (weidongxu-microsoft) merged 5 commits into
Azure:mainfrom
weidongxu-microsoft:copilot/refactor-management-autopr-evals
Aug 12, 2026
Merged

Refactor management AutoPR reviewer evals#50092
Weidong Xu (weidongxu-microsoft) merged 5 commits into
Azure:mainfrom
weidongxu-microsoft:copilot/refactor-management-autopr-evals

Conversation

@weidongxu-microsoft

@weidongxu-microsoft Weidong Xu (weidongxu-microsoft) commented Aug 11, 2026

Copy link
Copy Markdown
Member

Summary

  • move management AutoPR reviewer evals beside the management-autopr-review skill
  • rename suites to findings.eval.yaml, orchestration.eval.yaml, and true-negatives.eval.yaml
  • add an anti-trigger-only trigger.eval.yaml for generic PR and code-review requests
  • update Vally configuration, fixture paths, and local runner documentation

The skill is imported explicitly by its GitHub agentic workflow. The trigger suite therefore verifies only that interactive, general-purpose review requests do not invoke this specialized skill.

Verification

  • Vally lint: passed
  • trigger eval: 4/4 passed, 100%
  • findings eval: 95.2%
  • orchestration eval: 100%
  • true-negatives eval: 100%

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

Copy link
Copy Markdown
Azure Pipelines:
Successfully started running 1 pipeline(s).
34 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@weidongxu-microsoft
Weidong Xu (weidongxu-microsoft) marked this pull request as ready for review August 11, 2026 13:19
Copilot AI lite review requested due to automatic review settings August 11, 2026 13:19
@azure-pipelines

Copy link
Copy Markdown
Azure Pipelines:
Successfully started running 1 pipeline(s).
34 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Refactors the management AutoPR reviewer evaluation setup by relocating the eval suites alongside the management-autopr-review skill, renaming suites, adding an anti-trigger suite, and updating local runner/supporting configuration to match the new layout.

Changes:

  • Moved and renamed the management AutoPR reviewer eval suites under .github/skills/management-autopr-review/evals/ and updated fixture paths accordingly.
  • Added trigger.eval.yaml to ensure generic/interactive review prompts do not invoke the workflow-only management-autopr-review skill.
  • Updated the skill’s output contract example to code-format only the concern ID and added local eval runner + Vally config.

Reviewed changes

Copilot reviewed 9 out of 33 changed files in this pull request and generated no comments.

Show a summary per file
File Description
.github/skills/management-autopr-review/SKILL.md Tightens output formatting guidance (only code-format concern IDs).
.github/skills/management-autopr-review/evals/.vally.yaml New Vally suite configuration colocated with the skill.
.github/skills/management-autopr-review/evals/findings.eval.yaml Updated suite/fixture paths and added graders asserting concern-ID-only code formatting.
.github/skills/management-autopr-review/evals/orchestration.eval.yaml Updated suite/fixture/agent file paths to match new eval location.
.github/skills/management-autopr-review/evals/true-negatives.eval.yaml Updated suite/fixture paths to match new eval location.
.github/skills/management-autopr-review/evals/trigger.eval.yaml New anti-trigger capability eval for skill routing exclusions.
.github/skills/management-autopr-review/evals/run-evals.ps1 New local runner script for executing suites and validating workflow bot allowlist.
.github/skills/management-autopr-review/evals/README.md Updated local runner documentation to the new location.
.github/skills/management-autopr-review/evals/fixtures/positive-api-version-changed.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-api-version-metadata-map.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-api-version-overlap.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-breaking.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-folder.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-lro.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-manager-casing.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-manager-spelling.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-manager-unrelated.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-missing-release-plan.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-new-module.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-service-folder-mismatch.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-version-changelog-map.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/positive-version.md New synthetic fixture supporting findings coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-additive-beta-update.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-beta-breaking-changelog.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-generated-only-java.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-manager-clean.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-manager-legacy.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-manager-reordered.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-prior-concern.md New synthetic fixture supporting true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-prompt-injection.md New synthetic fixture supporting prompt-injection true-negative coverage.
.github/skills/management-autopr-review/evals/fixtures/tn-service-folder-match.md New synthetic fixture supporting true-negative coverage.
.github/skills/evals/management-autopr-reviewer/.vally.yaml Removed old eval configuration after relocation.
.github/skills/evals/management-autopr-reviewer/.gitignore Removed old ignore file after relocation.

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

@weidongxu-microsoft
Weidong Xu (weidongxu-microsoft) merged commit 382229a into Azure:main Aug 12, 2026
27 of 28 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