microsoft/download-profile-trace
Download a profiler trace file from Application Insights Profiler. Use this when asked to download, save, or export a profiler trace or .etl/.nettrace file.
npx skills add https://github.com/microsoft/code-optimizations-skills --skill download-profile-trace
This skill downloads a raw profiler trace artifact from the Application Insights Profiler dataplane API. The downloaded file can be opened in PerfView, Visual Studio, or other trace analysis tools.
az login)Follow the Standard Skill Preamble to check for existing investigation context and gather inputs.
In addition to the standard App ID, this skill requires:
P7D).Run the script in get-access-token.md to acquire a Bearer token for the profiler dataplane. See the token freshness and session scoping guidance in that document — re-acquire the token in the same command block as each API call.
Run the script in list-profile-traces.md to call the ingested artifacts endpoint and list available profiler traces.
Display only the 10 most recent traces by default. If more traces exist, tell the user the total count and offer to show more. The API can return hundreds of traces for active applications, which produces overwhelming output.
Present the results to the user with:
triggerTime — when the profiling session was triggered)When constructing the output filename, derive the file extension from two fields:
format → base extension: Etl → .etl, Nettrace → .nettrace, Netperf → .netperfblobUri → compression: if the URI ends in .zip, append .zip to the base extension> ⚠️ Do not use the blobUri extension as-is for the filename. The blobUri may show .etl.zip even when format is Nettrace, which would produce a misleading filename. Always use format for the base extension and blobUri only to detect .zip compression.
Include the role name in the filename to avoid collisions when downloading traces from different resources into the same directory, e.g., trace-myapp-2026-03-20T214135.nettrace.zip.
Let the user pick which trace to download. If there's only one result, confirm it with the user before proceeding.
Choose the download method based on whether the selected trace has an artifactId:
artifactId is available (not null)Run the script in download-trace.md with the selected artifactId. The script:
GET /artifacts/{artifactId})artifactId is nullFirst, run the script in resolve-trace-identifiers.md to query Application Insights customEvents for the ServiceProfilerSample event matching the selected trace. This can resolve:
ServiceProfilerContext) — if found, use the standard download in step 5aServiceProfilerContent) — if found, use the trace location download belowIf an artifact ID is resolved, go back to step 5a and download using that artifact ID.
If only a trace location ID is resolved, use download-trace-by-location.md to download. The trace location ID is a pipe-delimited v1 string:
v1|{stampId}|{appId}|{machineName}|{processId}|{etlFileSessionId}
The script:
POST /artifacts/byArtifactLocation?t={traceLocationId} to get a SAS-protected download URLIf neither identifier is resolved from the query, ask the user for the trace location ID or artifact ID manually.
After downloading, tell the user:
.zip, remind the user to extract it before analysisSee trace-file-info.md for information about trace file formats and how to open them.
Take microsoft/download-profile-trace from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.