Fix duplicate entries in Skills Loaded (and Agents Invoked) columns of evaluation table (#462)

* Initial plan * Fix duplicate entries in Skills Loaded column of evaluation table Agent-Logs-Url: https://github.com/dotnet/skills/sessions/2430176d-bf62-461d-bb79-781e34d7d084 Co-authored-by: danmoseley <6385855+danmoseley@users.noreply.github.com> * Remove accidentally committed .nuget binary Agent-Logs-Url: https://github.com/dotnet/skills/sessions/2430176d-bf62-461d-bb79-781e34d7d084 Co-authored-by: danmoseley <6385855+danmoseley@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: danmoseley <6385855+danmoseley@users.noreply.github.com>
2026-09-20 09:49:54 +08:00 · 2026-03-30 10:56:58 +00:00
parent 6248ef2b4b
commit d8d3f3d369
3 changed files with 105 additions and 2 deletions
@@ -417,3 +417,4 @@ FodyWeavers.xsd
 *.msm
 *.msp
 .skill-validator-results/
+.nuget/
@@ -523,7 +523,8 @@ public static class Reporter
                if (anyPluginRun && s.SkillActivationPlugin is { } saPlug)
                {
                    string plugActivation = FormatActivationCell(saPlug, s.ExpectActivation);
-                    skillsCol += $" / {plugActivation}";
+                    if (plugActivation != skillsCol)
+                        skillsCol += $" / {plugActivation}";
                }

                // Agents invoked column — show subagent activations
@@ -531,7 +532,8 @@ public static class Reporter
                if (anyPluginRun && s.SubagentActivationPlugin is { } saPlugAgent)
                {
                    string plugAgents = FormatSubagentCell(saPlugAgent);
-                    agentsCol += $" / {plugAgents}";
+                    if (plugAgents != agentsCol)
+                        agentsCol += $" / {plugAgents}";
                }

                var footnote = BuildVerdictFootnote(s, qualityDelta);
@@ -714,6 +714,106 @@ public class OverfittingJudgeTests
        Assert.DoesNotContain("### ❌ Skill validation errors", md);
    }

+    [Fact]
+    public void MarkdownTable_SkillsLoaded_NoDuplicateWhenIsolatedAndPluginIdentical()
+    {
+        var activation = new SkillActivationInfo(
+            Activated: true,
+            DetectedSkills: new List<string> { "my-skill" },
+            ExtraTools: new List<string> { "report_intent", "skill" },
+            SkillEventCount: 1);
+
+        var verdicts = new List<SkillVerdict>
+        {
+            new()
+            {
+                SkillName = "my-skill",
+                SkillPath = "/test",
+                Passed = true,
+                Scenarios = new List<ScenarioComparison>
+                {
+                    new()
+                    {
+                        ScenarioName = "sc1",
+                        Baseline = new RunResult(
+                            new RunMetrics { AgentOutput = "baseline" },
+                            new JudgeResult(new List<RubricScore>(), 3.5, "OK")),
+                        SkilledIsolated = new RunResult(
+                            new RunMetrics { AgentOutput = "skilled" },
+                            new JudgeResult(new List<RubricScore>(), 4.5, "Good")),
+                        SkilledPlugin = new RunResult(
+                            new RunMetrics { AgentOutput = "skilled-plugin" },
+                            new JudgeResult(new List<RubricScore>(), 4.5, "Good")),
+                        ImprovementScore = 0.25,
+                        Breakdown = new MetricBreakdown(0, 0, 0, 0, 0, 0, 0),
+                        SkillActivationIsolated = activation,
+                        SkillActivationPlugin = activation,
+                    }
+                },
+                OverallImprovementScore = 0.25,
+                Reason = "Pass",
+            }
+        };
+
+        var md = Reporter.GenerateMarkdownSummary(verdicts);
+
+        // Skills loaded cell should contain the activation info exactly once (no " / " separator)
+        Assert.DoesNotContain("my-skill; tools: report_intent, skill / ✅ my-skill", md);
+        Assert.Contains("✅ my-skill; tools: report_intent, skill", md);
+    }
+
+    [Fact]
+    public void MarkdownTable_SkillsLoaded_ShowsBothWhenIsolatedAndPluginDiffer()
+    {
+        var isolatedActivation = new SkillActivationInfo(
+            Activated: true,
+            DetectedSkills: new List<string> { "my-skill" },
+            ExtraTools: new List<string> { "report_intent" },
+            SkillEventCount: 1);
+        var pluginActivation = new SkillActivationInfo(
+            Activated: true,
+            DetectedSkills: new List<string> { "my-skill" },
+            ExtraTools: new List<string> { "report_intent", "skill" },
+            SkillEventCount: 1);
+
+        var verdicts = new List<SkillVerdict>
+        {
+            new()
+            {
+                SkillName = "my-skill",
+                SkillPath = "/test",
+                Passed = true,
+                Scenarios = new List<ScenarioComparison>
+                {
+                    new()
+                    {
+                        ScenarioName = "sc1",
+                        Baseline = new RunResult(
+                            new RunMetrics { AgentOutput = "baseline" },
+                            new JudgeResult(new List<RubricScore>(), 3.5, "OK")),
+                        SkilledIsolated = new RunResult(
+                            new RunMetrics { AgentOutput = "skilled" },
+                            new JudgeResult(new List<RubricScore>(), 4.5, "Good")),
+                        SkilledPlugin = new RunResult(
+                            new RunMetrics { AgentOutput = "skilled-plugin" },
+                            new JudgeResult(new List<RubricScore>(), 4.5, "Good")),
+                        ImprovementScore = 0.25,
+                        Breakdown = new MetricBreakdown(0, 0, 0, 0, 0, 0, 0),
+                        SkillActivationIsolated = isolatedActivation,
+                        SkillActivationPlugin = pluginActivation,
+                    }
+                },
+                OverallImprovementScore = 0.25,
+                Reason = "Pass",
+            }
+        };
+
+        var md = Reporter.GenerateMarkdownSummary(verdicts);
+
+        // Both activation entries should appear separated by " / "
+        Assert.Contains("✅ my-skill; tools: report_intent / ✅ my-skill; tools: report_intent, skill", md);
+    }
+
    // --- Prompt overfitting detection tests ---

    [Fact]