Add binlog MCP usage to other msbuild skills/agents

[JanKrivanek]

Followup of #673

Using binlog mcp in other msbuild skills/agents

[JanKrivanek]

/evaluate

[Copilot]

Pull request overview

Extends the dotnet-msbuild plugin’s existing binlog MCP integration by updating additional MSBuild skills and the build performance agent to prefer structured MCP binlog analysis (with a text-log replay fallback).

Changes:

• Added “Primary: binlog MCP (preferred)” workflows to multiple MSBuild skills (incremental build, evaluation perf, parallelism, bin/obj clash detection, ResolveProjectReferences guidance).
• Updated skill/agent frontmatter descriptions to mention binlog MCP usage and fallback behaviors.
• Updated build-perf agent workflow to use MCP tools for top-down performance analysis, with a text-log replay alternative.

Show a summary per file

File	Description
plugins/dotnet-msbuild/skills/resolve-project-references/SKILL.md	Adds MCP-preferred guidance for task self-time analysis when ResolveProjectReferences looks expensive.
plugins/dotnet-msbuild/skills/incremental-build/SKILL.md	Introduces an MCP-first workflow for diagnosing non-incremental rebuild causes.
plugins/dotnet-msbuild/skills/eval-performance/SKILL.md	Adds an MCP-first workflow for evaluation timing/property/import analysis.
plugins/dotnet-msbuild/skills/check-bin-obj-clash/SKILL.md	Reworks the primary workflow to use MCP tools to detect output/intermediate path clashes.
plugins/dotnet-msbuild/skills/build-parallelism/SKILL.md	Adds MCP-first steps for identifying expensive projects/targets and target timing.
plugins/dotnet-msbuild/agents/build-perf.agent.md	Switches the main top-down analysis flow to MCP tools, keeping a replay+grep alternate path.

Copilot's findings

Tip

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

Comments suppressed due to low confidence (1)

plugins/dotnet-msbuild/skills/resolve-project-references/SKILL.md:51

• In the fallback command, -fl and -flp are formatted in a confusing way (-fl "-flp:..."). This reads like -fl is taking an argument and differs from the syntax used elsewhere in this plugin. Consider using the standard form (-fl followed by -flp:..., with PowerShell-appropriate quoting if needed) to avoid copy/paste failures.

```bash
dotnet msbuild build.binlog -noconlog -fl "-flp:v=diag;logfile=full.log;performancesummary"
grep "Task Performance Summary" -A 50 full.log

• Files reviewed: 6/6 changed files
• Comments generated: 2

[github-actions]

Skill Validation Results

Skill	Scenario	Quality	Skills Loaded	Overfit	Verdict
resolve-project-references	Explain misleading ResolveProjectReferences time	3.3/5 → 5.0/5 🟢	✅ resolve-project-references; tools: skill, glob / ✅ resolve-project-references; tools: skill	✅ 0.16	✅
build-parallelism	Analyze build parallelism bottlenecks	3.7/5 → 4.3/5 🟢	✅ build-parallelism; tools: skill, binlog-binlog_overview, binlog-binlog_expensive_projects, binlog-binlog_projects, binlog-binlog_errors, glob, edit / ⚠️ NOT ACTIVATED	✅ 0.15	✅ [1]
incremental-build	Analyze incremental build issues	3.0/5 → 4.3/5 🟢	✅ incremental-build; tools: skill, bash / ⚠️ NOT ACTIVATED	✅ 0.16	✅ [2]
check-bin-obj-clash	Diagnose bin/obj output path clashes	4.3/5 → 5.0/5 🟢	✅ check-bin-obj-clash; tools: skill, glob, binlog-binlog_overview, binlog-binlog_projects, binlog-binlog_evaluations, binlog-binlog_double_writes, binlog-binlog_evaluation_properties, binlog-binlog_evaluation_global_properties / ✅ check-bin-obj-clash; tools: skill, binlog-binlog_overview, binlog-binlog_double_writes, binlog-binlog_evaluations, binlog-binlog_properties, binlog-binlog_evaluation_properties, binlog-binlog_diagnose	✅ 0.14	❌ [3]
eval-performance	Analyze MSBuild evaluation performance issues	5.0/5 → 4.7/5 🔴	✅ eval-performance; tools: skill	✅ 0.20	❌ [4]

[1] ⚠️ High run-to-run variance (CV=62%) — consider re-running with --runs 5
[2] ⚠️ High run-to-run variance (CV=88%) — consider re-running with --runs 5
[3] ⚠️ High run-to-run variance (CV=515%) — consider re-running with --runs 5. (Isolated) Quality improved but weighted score is -17.2% due to: judgment, tokens (91656 → 276424), tool calls (12 → 27), time (75.8s → 101.9s)
[4] (Plugin) Quality unchanged but weighted score is -9.4% due to: quality, tokens (41594 → 73529)

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

🔍 Full Results - additional metrics and failure investigation steps

▶ Sessions Visualisation -- interactive replay of all evaluation sessions

[JanKrivanek]

/evaluate

[Copilot]

Copilot's findings

• Files reviewed: 6/6 changed files
• Comments generated: 6

[github-actions]

Skill Validation Results

Skill	Scenario	Quality	Skills Loaded	Overfit	Verdict
resolve-project-references	Explain misleading ResolveProjectReferences time	3.3/5 → 5.0/5 🟢	✅ resolve-project-references; tools: skill	✅ 0.12	✅
build-parallelism	Analyze build parallelism bottlenecks	3.7/5 → 3.7/5 ⏰	✅ build-parallelism; tools: skill, binlog-binlog_overview, binlog-binlog_expensive_projects, binlog-binlog_errors, binlog-binlog_projects, edit, binlog-binlog_expensive_targets, glob / ⚠️ NOT ACTIVATED	✅ 0.20	❌ [1]
incremental-build	Analyze incremental build issues	3.0/5 → 4.7/5 🟢	✅ incremental-build; tools: skill, bash / ⚠️ NOT ACTIVATED	✅ 0.14	✅ [2]
check-bin-obj-clash	Diagnose bin/obj output path clashes	3.7/5 → 5.0/5 🟢	✅ check-bin-obj-clash; tools: glob, skill, binlog-binlog_overview, binlog-binlog_projects, binlog-binlog_double_writes, binlog-binlog_evaluations, binlog-binlog_evaluation_properties, binlog-binlog_evaluation_global_properties / ✅ check-bin-obj-clash; tools: glob, binlog-binlog_double_writes, binlog-binlog_diagnose, skill, binlog-binlog_evaluations, binlog-binlog_evaluation_properties, binlog-binlog_evaluation_global_properties, binlog-binlog_properties	✅ 0.14	❌ [3]
eval-performance	Analyze MSBuild evaluation performance issues	5.0/5 → 5.0/5	✅ eval-performance; tools: skill	✅ 0.17	❌ [4]

[1] ⚠️ High run-to-run variance (CV=432%) — consider re-running with --runs 5. (Isolated) Quality unchanged but weighted score is -7.1% due to: tokens (198097 → 381745), tool calls (15 → 25), time (82.4s → 107.6s)
[2] ⚠️ High run-to-run variance (CV=893%) — consider re-running with --runs 5
[3] ⚠️ High run-to-run variance (CV=62%) — consider re-running with --runs 5. (Isolated) Quality improved but weighted score is -23.5% due to: judgment, tokens (102832 → 220260), quality, tool calls (12 → 23), time (68.7s → 95.4s)
[4] ⚠️ High run-to-run variance (CV=50%) — consider re-running with --runs 5

⏰ timeout — run(s) hit the (160s) scenario timeout limit; scoring may be impacted by aborting model execution before it could produce its full output (increase via timeout in eval.yaml)

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

🔍 Full Results - additional metrics and failure investigation steps

▶ Sessions Visualisation -- interactive replay of all evaluation sessions