Upgrade agentic workflows and fix stale PR cleanup

[Evangelink]

Summary

This PR combines an agentic workflow runtime upgrade and stale-PR cleanup fix with related dotnet-test skill and evaluation improvements already present on the branch.

Agentic workflows

• upgrades the repository's gh-aw runtime, dispatcher, action pins, maintenance workflow, and generated lock files
• fixes stale PR cleanup so it searches only PRs older than the cutoff, paginates the complete result set, includes drafts, and processes the oldest closures before warnings
• adds an explicit noop path when no stale PR action is required
• adds the gh-aw router skill and Copilot agent setup workflow generated for the upgraded integration

The failed stale-PR run emitted only one warning for PR 703 because it inspected only the newest 100 open pull requests. The corrected search surfaces all eight old draft PRs that were previously beyond the first page.

dotnet-test skills and evaluations

• clarifies the Research-Plan-Implement handoffs and language-specific extension guidance used by the code-testing agents
• improves MSTest and xUnit-to-MSTest migration guidance and adds a mapping reference
• adds and updates eval coverage for untested-source discovery, grading, and MSTest migration, including their fixture projects

Why copilot-setup-steps.yml is new

gh aw upgrade generated this repository-level setup workflow so GitHub Copilot coding agent sessions can install the matching gh-aw CLI before working with agentic workflows. GitHub recognizes the copilot-setup-steps job name and runs it during Copilot agent environment setup. The file is part of the upgraded gh-aw integration, not part of the stale PR cleanup runtime itself.

Security review

The post-merge gh-aw compilation flagged COPILOT_PAT_0 through COPILOT_PAT_9 as newly referenced restricted secrets. These references come from main's shared PAT-pool integration. I reviewed the flow: the trusted pat_pool job only checks which secrets are non-empty and emits a numeric index; the agent job then resolves that index to the matching secret through the case() expression. Secret values are not emitted by the selector or exposed to untrusted steps. No new actions or redirect changes were flagged by compilation. The secret additions are safe to approve.

[Copilot]

Pull request overview

This PR upgrades the repository’s gh-aw/agentic workflow runtime and regenerates the associated lock files, while also fixing the stale PR cleanup agent so it reliably finds and processes all old open PRs (including drafts) by paginating through the full result set and providing a noop path when nothing needs action.

Changes:

• Upgraded gh-aw runtime/action pins and regenerated multiple .lock.yml workflows plus the generated maintenance workflow.
• Updated the stale PR agent instructions to search only older PRs, include drafts, paginate all pages, and process oldest activity first.
• Added a Copilot Agent “setup steps” workflow and introduced an “agentic-workflows” router skill/agent documentation.

Show a summary per file

File	Description
.github/workflows/markdown-linter.md	Updates action pins (checkout + super-linter).
.github/workflows/devops-health-investigate.lock.yml	Regenerated lock workflow for newer gh-aw runtime and updated safe outputs / MCP gateway wiring.
.github/workflows/copilot-setup-steps.yml	Adds a dedicated Copilot Agent setup workflow (checkout + gh-aw CLI install).
.github/workflows/close-stale-prs.agent.md	Updates agent instructions to paginate and include drafts; adds noop behavior.
.github/workflows/close-stale-prs.agent.lock.yml	Regenerated lock workflow for the stale PR agent with updated runtime and tooling.
.github/workflows/agentics-maintenance.yml	Regenerated maintenance workflow for v0.82.8 with split expiration cleanup jobs and updated caching/forecast steps.
.github/skills/agentic-workflows/SKILL.md	Adds an agentic workflow “router” skill for selecting the right gh-aw prompt/guide.
.github/aw/actions-lock.json	Updates locked action versions (notably checkout v7 and gh-aw actions v0.82.8).
.github/agents/agentic-workflows.md	Updates dispatcher agent metadata and points prompt URLs to raw.githubusercontent.com main branch references.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 14/15 changed files
• Comments generated: 2

[Copilot]

Copilot's findings

• Files reviewed: 14/15 changed files
• Comments generated: 2

[Copilot]

Copilot's findings

• Files reviewed: 34/35 changed files
• Comments generated: 2

[Copilot]

Copilot's findings

• Files reviewed: 34/35 changed files
• Comments generated: 0 new

[github-actions]

👋 @Evangelink — this PR has merge conflict. When you're ready, please address the feedback and push an update; the triage bot will pick up the next state automatically. (Add the no-stale label to silence further pings.)

[github-actions]

Skill Coverage Report

	Plugin	Skill	Covered	Coverage
✅	dotnet-test-migration	migrate-mstest-v1v2-to-v3	15/16	93.8%
❌	dotnet-test-migration	migrate-xunit-to-mstest	0/1	0%
✅	dotnet-test	code-testing-agent	3/3	100%
✅	dotnet-test	grade-tests	28/28	100%

Uncovered: dotnet-test-migration/migrate-mstest-v1v2-to-v3

• [CodePattern] Assert.AreNotEqual (line 147)

Uncovered: dotnet-test-migration/migrate-xunit-to-mstest

• [CodePattern] Parallelize (line 106)

[Evangelink]

/evaluate

[Copilot]

Copilot's findings

• Files reviewed: 42/42 changed files
• Comments generated: 1

[github-actions]

Skill Validation Results

Skill	Scenario	Quality	Skills Loaded	Overfit	Verdict
migrate-mstest-v1v2-to-v3	Migrate MSTest v1 project with assembly reference	3.0/5 → 4.7/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Migrate MSTest v2 NuGet project to v3	3.0/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Fix Assert.AreEqual object overload errors after v3 upgrade	4.0/5 → 4.3/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill, bash	✅ 0.07	✅ [1]
migrate-mstest-v1v2-to-v3	Migrate from .testsettings to .runsettings	4.0/5 → 4.0/5	✅ migrate-mstest-v1v2-to-v3; tools: skill, bash	✅ 0.07	✅ [2]
migrate-mstest-v1v2-to-v3	Fix DataRow type mismatch errors after v3 upgrade	3.0/5 → 2.7/5 🔴	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	❌ [3]
migrate-mstest-v1v2-to-v3	Migrate to MSTest.Sdk project style	3.3/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill, bash	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Handle dropped target framework during v3 migration	2.0/5 → 4.7/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Migrate complex MSTest v2 project with testsettings, DataRow issues, and dropped TFM	4.3/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅ [4]
migrate-mstest-v1v2-to-v3	Correctly identify MSTest v1 vs v2 and recommend different migration paths	4.7/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill, glob / ✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Execute and verify MSTest v2 to v3 migration	4.3/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅ [5]
migrate-xunit-to-mstest	Migrate basic xUnit v2 project to MSTest v4 preserving VSTest	4.0/5 → 4.3/5 🟢	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.11	✅ [6]
migrate-xunit-to-mstest	Migrate basic xUnit v3 project to MSTest v4 preserving VSTest	4.0/5 → 1.7/5 🔴	✅ migrate-xunit-to-mstest; tools: skill, read_bash	✅ 0.11	❌ [7]
migrate-xunit-to-mstest	Stop when target framework is unsupported by MSTest v4	1.7/5 → 2.0/5 🟢	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, glob	✅ 0.11	✅ [8]
migrate-xunit-to-mstest	Convert IClassFixture to ClassInitialize	2.7/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, read_bash, grep	✅ 0.11	❌
migrate-xunit-to-mstest	Handle ICollectionFixture explicitly (do not silently widen scope)	3.3/5 → 2.7/5 🔴	✅ migrate-xunit-to-mstest; tools: skill, read_bash / ✅ migrate-xunit-to-mstest; tools: skill, grep, read_bash	✅ 0.11	❌ [9]
migrate-xunit-to-mstest	Convert ITestOutputHelper to TestContext	4.7/5 → 5.0/5 🟢	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, grep	✅ 0.11	❌ [10]
migrate-xunit-to-mstest	Map exception assertions correctly (xUnit Throws -> MSTest ThrowsExactly)	4.3/5 → 4.3/5	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.11	❌ [11]
migrate-xunit-to-mstest	Convert MemberData and TheoryData to DynamicData	3.3/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.11	❌
migrate-xunit-to-mstest	Convert Skip, Trait, and Timeout	2.3/5 → 2.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, read_bash	✅ 0.11	❌ [12]
migrate-xunit-to-mstest	Preserve xUnit parallelization default with [assembly: Parallelize]	2.7/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.11	❌
migrate-xunit-to-mstest	Convert xUnit v3 TestContext.Current.CancellationToken	4.0/5 → 4.3/5 🟢	✅ migrate-xunit-to-mstest; tools: skill, grep / ✅ migrate-xunit-to-mstest; tools: skill, grep, read_bash	✅ 0.11	✅ [13]
migrate-xunit-to-mstest	Recognize project already on MSTest	5.0/5 → 5.0/5	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.11	❌ [14]
code-testing-agent	Generate tests for ContosoUniversity ASP.NET Core MVC app	3.0/5 → 3.0/5	✅ code-testing-agent; tools: skill / ✅ code-testing-agent; tools: skill, task, read_agent	🟡 0.21	❌ [15]
code-testing-agent	Generate pytest tests for the Flask tasks API (Python polyglot)	4.3/5 → 4.0/5 🔴	✅ code-testing-agent; tools: skill, edit / ✅ code-testing-agent; tools: skill	🟡 0.21	❌ [16]
code-testing-agent	Generate Vitest tests for the shopping-cart library (TypeScript polyglot)	5.0/5 → 4.7/5 🔴	✅ code-testing-agent; tools: skill, edit	🟡 0.21	❌
code-testing-agent	Does not revert a gutted-looking workspace (workspace integrity)	5.0/5 → 5.0/5	⚠️ NOT ACTIVATED	🟡 0.21	❌ [17]
find-untested-sources	Identify unpaired C# source files and suggest a test location	4.0/5 → 4.0/5	⚠️ NOT ACTIVATED	🟡 0.23	❌ [18]
find-untested-sources	Reject non-C# source discovery	4.0/5 → 4.0/5	ℹ️ not activated (expected)	🟡 0.23	❌ [19]
grade-tests	Grade a curated list of MSTest tests with mixed quality	4.0/5 → 4.7/5 🟢	✅ grade-tests; tools: skill, glob / ✅ grade-tests; tools: skill, bash, glob	🟡 0.21	✅ [20]
grade-tests	Grade pytest test methods using the same rubric	5.0/5 → 4.3/5 🔴	✅ grade-tests; tools: skill / ✅ grade-tests; tools: skill, glob, bash, stop_bash	🟡 0.21	✅ [21]
grade-tests	Decline to grade the entire workspace when no test list is provided	1.0/5 → 4.7/5 🟢	✅ grade-tests; tools: skill / ✅ grade-tests; test-analysis-extensions; tools: skill	🟡 0.21	❌ [22]
grade-tests	Grade Go table-driven tests without misreading the loop as branching	4.7/5 → 3.7/5 🔴	✅ grade-tests; tools: skill, glob / ✅ grade-tests; tools: skill, bash, glob	🟡 0.21	❌ [23]
grade-tests	Grade tests when the production code under test is unavailable	4.0/5 → 4.0/5	✅ grade-tests; tools: skill, glob / ✅ grade-tests; tools: skill, bash, glob	🟡 0.21	❌ [24]

[1] ⚠️ High run-to-run variance (CV=5742%) — consider re-running with --runs 5
[2] ⚠️ High run-to-run variance (CV=6045%) — consider re-running with --runs 5
[3] ⚠️ High run-to-run variance (CV=290%) — consider re-running with --runs 5
[4] ⚠️ High run-to-run variance (CV=160%) — consider re-running with --runs 5
[5] ⚠️ High run-to-run variance (CV=120%) — consider re-running with --runs 5
[6] ⚠️ High run-to-run variance (CV=923%) — consider re-running with --runs 5
[7] ⚠️ High run-to-run variance (CV=80%) — consider re-running with --runs 5
[8] ⚠️ High run-to-run variance (CV=153%) — consider re-running with --runs 5
[9] ⚠️ High run-to-run variance (CV=404%) — consider re-running with --runs 5
[10] (Isolated) Quality improved but weighted score is -4.7% due to: tokens (68870 → 112259), time (33.3s → 45.6s), tool calls (8 → 10)
[11] (Isolated) Quality unchanged but weighted score is -4.9% due to: tokens (63565 → 112950), tool calls (8 → 11)
[12] ⚠️ High run-to-run variance (CV=260%) — consider re-running with --runs 5
[13] ⚠️ High run-to-run variance (CV=216%) — consider re-running with --runs 5
[14] ⚠️ High run-to-run variance (CV=91%) — consider re-running with --runs 5. (Plugin) Quality unchanged but weighted score is -5.9% due to: tokens (25505 → 47685), tool calls (3 → 4), time (10.2s → 13.0s)
[15] ⚠️ High run-to-run variance (CV=84%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -12.8% due to: judgment, tokens (1362383 → 1536268)
[16] ⚠️ High run-to-run variance (CV=135%) — consider re-running with --runs 5
[17] ⚠️ High run-to-run variance (CV=53%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -16.7% due to: judgment, quality, time (44.2s → 55.3s)
[18] ⚠️ High run-to-run variance (CV=61%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -4.0% due to: quality
[19] (Isolated) Quality unchanged but weighted score is -2.2% due to: quality
[20] ⚠️ High run-to-run variance (CV=137%) — consider re-running with --runs 5
[21] (Plugin) Quality dropped but weighted score is +2.0% due to: completion (✗ → ✓)
[22] (Plugin) Quality unchanged but weighted score is -5.2% due to: tokens (58137 → 97476), time (27.1s → 47.1s)
[23] ⚠️ High run-to-run variance (CV=91%) — consider re-running with --runs 5
[24] ⚠️ High run-to-run variance (CV=68%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -24.1% due to: judgment, tokens (26456 → 54585), quality, tool calls (2 → 4), time (16.9s → 30.0s)

Model: claude-opus-4.6 | Judge: claude-opus-4.6

🔍 Full Results - additional metrics and failure investigation steps

To investigate failures, paste this to your AI coding agent:

For PR 916 in dotnet/skills, download eval artifacts with gh run download 29738308377 --repo dotnet/skills --pattern "skill-validator-results-*" --dir ./eval-results, then fetch https://raw.githubusercontent.com/dotnet/skills/3442f30f182e45981e369b43cde6c3b18a6a90f7/eng/skill-validator/src/docs/InvestigatingResults.md and follow it to analyze the results.json files. Diagnose each failure, suggest fixes to the eval.yaml and skill content, and tell me what to fix first.

▶ Sessions Visualisation -- interactive replay of all evaluation sessions
📊 Session Analytics (preview) -- aggregated metrics across evaluation sessions

[Copilot]

Copilot's findings

• Files reviewed: 42/42 changed files
• Comments generated: 0 new

[github-actions]

Skill Validation Results

Skill	Scenario	Quality	Skills Loaded	Overfit	Verdict
migrate-mstest-v1v2-to-v3	Migrate MSTest v1 project with assembly reference	3.0/5 → 4.7/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Migrate MSTest v2 NuGet project to v3	3.0/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Fix Assert.AreEqual object overload errors after v3 upgrade	4.0/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅ [1]
migrate-mstest-v1v2-to-v3	Migrate from .testsettings to .runsettings	4.0/5 → 4.0/5	✅ migrate-mstest-v1v2-to-v3; tools: skill, bash	✅ 0.07	✅ [2]
migrate-mstest-v1v2-to-v3	Fix DataRow type mismatch errors after v3 upgrade	3.0/5 → 3.0/5	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅ [3]
migrate-mstest-v1v2-to-v3	Migrate to MSTest.Sdk project style	3.3/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill, edit, bash	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Handle dropped target framework during v3 migration	2.0/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill, edit / ✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Migrate complex MSTest v2 project with testsettings, DataRow issues, and dropped TFM	4.7/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅ [4]
migrate-mstest-v1v2-to-v3	Correctly identify MSTest v1 vs v2 and recommend different migration paths	4.3/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill, glob / ✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-mstest-v1v2-to-v3	Execute and verify MSTest v2 to v3 migration	4.0/5 → 5.0/5 🟢	✅ migrate-mstest-v1v2-to-v3; tools: skill	✅ 0.07	✅
migrate-xunit-to-mstest	Migrate basic xUnit v2 project to MSTest v4 preserving VSTest	4.0/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, read_bash	✅ 0.08	❌
migrate-xunit-to-mstest	Migrate basic xUnit v3 project to MSTest v4 preserving VSTest	3.0/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill, bash / ✅ migrate-xunit-to-mstest; tools: skill, bash, read_bash	✅ 0.08	❌
migrate-xunit-to-mstest	Stop when target framework is unsupported by MSTest v4	1.0/5 → 1.7/5 🟢	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.08	✅ [5]
migrate-xunit-to-mstest	Convert IClassFixture to ClassInitialize	2.3/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill, bash / ✅ migrate-xunit-to-mstest; tools: skill, bash, grep, read_bash	✅ 0.08	❌
migrate-xunit-to-mstest	Handle ICollectionFixture explicitly (do not silently widen scope)	2.7/5 → 1.0/5 🔴	✅ migrate-xunit-to-mstest; tools: skill, bash / ✅ migrate-xunit-to-mstest; tools: skill, grep, bash, read_bash	✅ 0.08	❌
migrate-xunit-to-mstest	Convert ITestOutputHelper to TestContext	4.7/5 → 5.0/5 🟢	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, grep	✅ 0.08	✅ [6]
migrate-xunit-to-mstest	Map exception assertions correctly (xUnit Throws -> MSTest ThrowsExactly)	5.0/5 → 5.0/5	✅ migrate-xunit-to-mstest; tools: skill, grep / ✅ migrate-xunit-to-mstest; tools: skill	✅ 0.08	❌ [7]
migrate-xunit-to-mstest	Convert MemberData and TheoryData to DynamicData	5.0/5 → 5.0/5	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, read_bash	✅ 0.08	❌ [8]
migrate-xunit-to-mstest	Convert Skip, Trait, and Timeout	4.0/5 → 4.0/5	✅ migrate-xunit-to-mstest; tools: skill / ✅ migrate-xunit-to-mstest; tools: skill, grep, read_bash	✅ 0.08	❌ [9]
migrate-xunit-to-mstest	Preserve xUnit parallelization default with [assembly: Parallelize]	3.7/5 → 4.0/5 🟢	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.08	✅ [10]
migrate-xunit-to-mstest	Convert xUnit v3 TestContext.Current.CancellationToken	3.7/5 → 3.7/5	✅ migrate-xunit-to-mstest; tools: skill, grep	✅ 0.08	✅ [11]
migrate-xunit-to-mstest	Recognize project already on MSTest	5.0/5 → 5.0/5	✅ migrate-xunit-to-mstest; tools: skill	✅ 0.08	❌ [12]
code-testing-agent	Generate tests for ContosoUniversity ASP.NET Core MVC app	3.0/5 → 3.0/5	✅ code-testing-agent; tools: skill / ✅ code-testing-agent; tools: edit, skill, task, read_agent, glob	🟡 0.25	❌ [13]
code-testing-agent	Generate pytest tests for the Flask tasks API (Python polyglot)	5.0/5 → 4.3/5 🔴	✅ code-testing-agent; tools: skill, edit / ✅ code-testing-agent; tools: skill	🟡 0.25	❌ [14]
code-testing-agent	Generate Vitest tests for the shopping-cart library (TypeScript polyglot)	5.0/5 → 5.0/5	✅ code-testing-agent; tools: skill	🟡 0.25	❌ [15]
code-testing-agent	Does not revert a gutted-looking workspace (workspace integrity)	5.0/5 → 5.0/5	⚠️ NOT ACTIVATED	🟡 0.25	❌ [16]
find-untested-sources	Identify unpaired C# source files and suggest a test location	4.0/5 → 4.0/5	⚠️ NOT ACTIVATED	🟡 0.32	❌ [17]
find-untested-sources	Reject non-C# source discovery	3.3/5 → 3.3/5	ℹ️ not activated (expected)	🟡 0.32	❌ [18]
grade-tests	Grade a curated list of MSTest tests with mixed quality	4.0/5 → 5.0/5 🟢	✅ grade-tests; tools: skill, glob / ✅ grade-tests; tools: skill, glob, bash	✅ 0.20	✅ [19]
grade-tests	Grade pytest test methods using the same rubric	5.0/5 → 4.0/5 🔴	✅ grade-tests; tools: skill, glob / ✅ grade-tests; tools: skill, glob, bash	✅ 0.20	✅ [20]
grade-tests	Decline to grade the entire workspace when no test list is provided	1.0/5 → 2.0/5 🟢	✅ grade-tests; tools: skill, glob	✅ 0.20	❌ [21]
grade-tests	Grade Go table-driven tests without misreading the loop as branching	4.0/5 → 4.0/5	✅ grade-tests; tools: skill, glob	✅ 0.20	❌ [22]
grade-tests	Grade tests when the production code under test is unavailable	4.0/5 → 4.0/5	✅ grade-tests; tools: skill, glob / ✅ grade-tests; tools: skill, glob, bash	✅ 0.20	❌ [23]

[1] ⚠️ High run-to-run variance (CV=128%) — consider re-running with --runs 5
[2] ⚠️ High run-to-run variance (CV=2308%) — consider re-running with --runs 5
[3] ⚠️ High run-to-run variance (CV=309%) — consider re-running with --runs 5
[4] ⚠️ High run-to-run variance (CV=551%) — consider re-running with --runs 5
[5] ⚠️ High run-to-run variance (CV=57%) — consider re-running with --runs 5
[6] ⚠️ High run-to-run variance (CV=55%) — consider re-running with --runs 5
[7] (Plugin) Quality unchanged but weighted score is -7.3% due to: tokens (73203 → 202831), tool calls (8 → 13), time (93.8s → 122.1s)
[8] ⚠️ High run-to-run variance (CV=65%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -7.1% due to: tokens (69827 → 133285), time (70.6s → 115.9s), tool calls (8 → 11)
[9] ⚠️ High run-to-run variance (CV=226%) — consider re-running with --runs 5. (Plugin) Quality unchanged but weighted score is -10.0% due to: tokens (85396 → 325064), tool calls (9 → 19), time (88.6s → 179.0s)
[10] ⚠️ High run-to-run variance (CV=267%) — consider re-running with --runs 5
[11] ⚠️ High run-to-run variance (CV=8910%) — consider re-running with --runs 5
[12] ⚠️ High run-to-run variance (CV=82%) — consider re-running with --runs 5. (Plugin) Quality unchanged but weighted score is -3.3% due to: tokens (25519 → 47668), tool calls (3 → 4)
[13] ⚠️ High run-to-run variance (CV=64%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -15.2% due to: judgment, quality, tokens (993607 → 1249274)
[14] ⚠️ High run-to-run variance (CV=147%) — consider re-running with --runs 5
[15] (Isolated) Quality unchanged but weighted score is -3.8% due to: quality, tokens (222881 → 255931)
[16] ⚠️ High run-to-run variance (CV=103%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -4.2% due to: quality
[17] ⚠️ High run-to-run variance (CV=110%) — consider re-running with --runs 5. (Plugin) Quality unchanged but weighted score is -0.7% due to: tokens (61075 → 77121)
[18] (Isolated) Quality unchanged but weighted score is -12.9% due to: judgment, tokens (59444 → 67919)
[19] ⚠️ High run-to-run variance (CV=51%) — consider re-running with --runs 5
[20] ⚠️ High run-to-run variance (CV=611%) — consider re-running with --runs 5
[21] ⚠️ High run-to-run variance (CV=343%) — consider re-running with --runs 5. (Plugin) Quality unchanged but weighted score is -8.2% due to: tokens (52505 → 114630), time (24.7s → 49.6s), tool calls (7 → 9)
[22] ⚠️ High run-to-run variance (CV=851%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -9.8% due to: judgment, tokens (26722 → 64606), quality, time (18.0s → 33.4s), tool calls (3 → 5)
[23] (Plugin) Quality unchanged but weighted score is -9.4% due to: tokens (26459 → 133143), tool calls (2 → 6), time (19.8s → 35.0s)

Model: claude-opus-4.6 | Judge: claude-opus-4.6

🔍 Full Results - additional metrics and failure investigation steps

To investigate failures, paste this to your AI coding agent:

For PR 916 in dotnet/skills, download eval artifacts with gh run download 29744473853 --repo dotnet/skills --pattern "skill-validator-results-*" --dir ./eval-results, then fetch https://raw.githubusercontent.com/dotnet/skills/f69d27d16d27d71b47539a342eccb413e957badf/eng/skill-validator/src/docs/InvestigatingResults.md and follow it to analyze the results.json files. Diagnose each failure, suggest fixes to the eval.yaml and skill content, and tell me what to fix first.

▶ Sessions Visualisation -- interactive replay of all evaluation sessions
📊 Session Analytics (preview) -- aggregated metrics across evaluation sessions