Add aot-impact-analysis skill for AOT size and startup analysis - #55085
Draft
baronfel wants to merge 2 commits into
Draft
Add aot-impact-analysis skill for AOT size and startup analysis#55085baronfel wants to merge 2 commits into
baronfel wants to merge 2 commits into
Conversation
Adds a reusable agent skill that automates the binary-size and startup-performance analysis used when enabling NativeAOT functionality in the dotnet CLI. The skill follows the thin-core + subskill-pages structure: - SKILL.md: thin orchestrator (when to measure, the always-apply rules) - size-analysis.md: before/after dotnet-aot.mstat diff via sizoscope-cli - startup-and-telemetry.md: redist-muxer startup timing + Aspire Dashboard OTel - writing-the-summary.md: assembling the PR-ready impact summary - references/perf-and-otel.md: OTLP env-var contract + aspire otel surface - scripts/: Measure-AotSize.ps1, Measure-AotStartup.ps1, Start-AspireDashboard.ps1 Size methodology mirrors the aot-size-analysis CI workflow; startup is measured through the real redist dotnet.exe muxer (DOTNET_CLI_ENABLEAOT) rather than a standalone host. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Contributor
There was a problem hiding this comment.
Pull request overview
Adds a new reusable GitHub skill (aot-impact-analysis) under .github/skills/ to standardize how contributors measure NativeAOT size impact (via sizoscope-cli on dotnet-aot.mstat) and startup impact (timings through the redist muxer + optional OpenTelemetry capture in Aspire Dashboard) and then assemble a PR-ready summary.
Changes:
- Introduces the
aot-impact-analysisskill orchestrator (SKILL.md) with progressive-disclosure subpages for size, startup/telemetry, and summary writing. - Adds PowerShell scripts to automate size diffing (baseline worktree publish +
sizoscope-cli) and startup timing/telemetry export. - Adds reference documentation for the CLI’s OTLP/exporter environment variable contract and Aspire Dashboard usage.
Reviewed changes
Copilot reviewed 8 out of 8 changed files in this pull request and generated 7 comments.
Show a summary per file
| File | Description |
|---|---|
| .github/skills/aot-impact-analysis/SKILL.md | Skill entry point and “always apply” rules + how to run size/startup. |
| .github/skills/aot-impact-analysis/size-analysis.md | Explains size methodology and how to interpret sizoscope output. |
| .github/skills/aot-impact-analysis/startup-and-telemetry.md | Explains muxer-based startup measurement and Aspire telemetry capture. |
| .github/skills/aot-impact-analysis/writing-the-summary.md | Guidance for composing the PR-ready impact summary. |
| .github/skills/aot-impact-analysis/references/perf-and-otel.md | Documents OTLP env vars and aspire otel query surface. |
| .github/skills/aot-impact-analysis/scripts/Measure-AotSize.ps1 | Automates baseline/PR publish + .mstat diff + Markdown report output. |
| .github/skills/aot-impact-analysis/scripts/Measure-AotStartup.ps1 | Automates managed-vs-AOT timing + optional exporter-on trace capture. |
| .github/skills/aot-impact-analysis/scripts/Start-AspireDashboard.ps1 | Starts Aspire Dashboard via aspire CLI or container fallback. |
Comment on lines
+91
to
+101
| function Find-Native([string]$root, [string]$config, [string]$leaf) { | ||
| if ($leaf -eq 'dotnet-aot.dll') { | ||
| $base = Join-Path $root "artifacts/bin/dotnet-aot/$config" | ||
| } else { | ||
| $base = Join-Path $root "artifacts/obj/dotnet-aot/$config" | ||
| } | ||
| $hit = Get-ChildItem -Path $base -Recurse -Filter $leaf -ErrorAction SilentlyContinue | | ||
| Where-Object { $_.FullName -match 'native' } | Select-Object -First 1 | ||
| if (-not $hit) { throw "Could not find native/$leaf under $base." } | ||
| return $hit.FullName | ||
| } |
Comment on lines
+9
to
+13
| official container image (`mcr.microsoft.com/dotnet/aspire-dashboard`). Either way the dashboard | ||
| exposes: | ||
| * UI http://localhost:18888 | ||
| * OTLP/gRPC http://localhost:4317 (point the CLI's OTEL_EXPORTER_OTLP_ENDPOINT here) | ||
| * OTLP/HTTP http://localhost:4318 |
Comment on lines
+27
to
+31
| Endpoints: UI `http://localhost:18888`, OTLP/gRPC `http://localhost:4317`, | ||
| OTLP/HTTP `http://localhost:4318`. It prefers the `aspire` CLI and falls back to | ||
| the `mcr.microsoft.com/dotnet/aspire-dashboard` container. With the CLI it prints | ||
| a `…/login?t=<token>` URL — pass that whole URL to `aspire otel --dashboard-url`. | ||
|
|
Comment on lines
+35
to
+40
| `Start-AspireDashboard.ps1` runs a standalone dashboard. Default endpoints: | ||
|
|
||
| - UI: `http://localhost:18888` | ||
| - OTLP/gRPC: `http://localhost:4317` -> set `OTEL_EXPORTER_OTLP_ENDPOINT` to this | ||
| - OTLP/HTTP: `http://localhost:4318` | ||
|
|
Comment on lines
+20
to
+21
| Dashboard). All scripts are PowerShell, self-locate the repo root, and use the | ||
| repo-local SDK (`.dotnet/dotnet.exe`); run them from anywhere in the worktree. |
Comment on lines
+36
to
+39
| 2. Adds a **detached git worktree** at the baseline ref beside the repo and | ||
| publishes the same project there, using the current worktree's repo-local SDK | ||
| (`.dotnet/dotnet.exe`). The baseline restore pulls the baseline package set — | ||
| expect a cold restore on first run. |
Comment on lines
+101
to
+107
| function Invoke-Timed([string]$exe, [string]$argline, [hashtable]$env, [int]$count, [int]$warmup) { | ||
| $applied = @{} | ||
| foreach ($k in $env.Keys) { $applied[$k] = [Environment]::GetEnvironmentVariable($k); [Environment]::SetEnvironmentVariable($k, $env[$k]) } | ||
| try { | ||
| $argv = $argline.Split(' ', [StringSplitOptions]::RemoveEmptyEntries) | ||
| for ($i = 0; $i -lt $warmup; $i++) { & $exe @argv *> $null } | ||
| $times = New-Object System.Collections.Generic.List[double] |
Ports the three PowerShell helpers to bash for Linux/macOS shells, keeping the PowerShell versions canonical: - measure-aot-size.sh (<-> Measure-AotSize.ps1) - measure-aot-startup.sh (<-> Measure-AotStartup.ps1) - start-aspire-dashboard.sh (<-> Start-AspireDashboard.ps1) The bash ports mirror the same parameters as --kebab-case flags, write the same Markdown artifacts, and use the same methodology (sizoscope diff vs the fork point; redist-muxer startup timing with DOTNET_CLI_ENABLEAOT; OTLP export to a reachable Aspire Dashboard). Startup percentile math matches the PowerShell implementation; sub-second timing uses bash 5 EPOCHREALTIME or gdate. Scripts are marked executable and pass bash -n. SKILL.md and the subskill pages note the bash forms. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
mthalman
reviewed
Jul 1, 2026
Member
There was a problem hiding this comment.
This file, and the other non-SKILL.md files at the skill root directory, should probably be moved to the references directory as these are on-demand markdown files to supplement the main skill.
| "did startup regress", "profile the AOT entrypoint". | ||
| license: MIT | ||
| metadata: | ||
| portability: repo-local |
Member
Author
|
@mthalman I'm gonna mark this draft - this is waiting on some work from @JeremyKuhne and is pattered off of his jeremykuhne/agent-skills repo so will need more work before it aligns to the SDK repo patterns. |
baronfel
marked this pull request as draft
July 1, 2026 14:27
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a reusable agent skill,
aot-impact-analysis, that automates the binary-size and startup-performance analysis we run when enabling NativeAOT functionality in the dotnet CLI (e.g. #54970).Structure
Follows a thin-core + subskill-pages layout (modeled on the JeremyKuhne/agent-skills conventions): a small
SKILL.mdthat loads on every trigger, plus sibling pages loaded on demand.SKILL.mdsize-analysis.mddotnet-aot.mstatdiff viasizoscope-clistartup-and-telemetry.mdwriting-the-summary.mdreferences/perf-and-otel.mdaspire otelcommand surfacescripts/Measure-AotSize.ps1,Measure-AotStartup.ps1,Start-AspireDashboard.ps1Notes
aot-size-analysisCI workflow (sizoscope on the per-platformdotnet-aot.mstat, baselined at the fork point).dotnet.exemuxer (DOTNET_CLI_ENABLEAOT), not a standalone host, so it exercises the same dispatch users hit..dotnet/dotnet.exe);SKILL.mdvalidates against.github/skills/ValidateSkill.cs.