feat: add Foundry Loom experiment tracking skill

[HarshaVardhanBabu (babu-namburi)]

Summary

• add an experiment-tracking sub-skill to microsoft-foundry with concise azd ai loom routing triggers
• document all 18 Foundry Loom run inspection, trace/span query, OTLP ingestion, and W&B-compatible command surfaces from Add azure.ai.loom CLI support for Foundry experiment tracking APIs Azure/azure-dev#9796
• document azure.ai.loom installation, the azd >= 1.32.0 requirement, endpoint resolution, authentication, JSON preservation, and result limits
• add Vally routing and command-selection coverage for azd ai loom run

How to use this skill

experiment-tracking is a sub-skill of microsoft-foundry; users do not invoke it by a separate skill name. Copilot loads microsoft-foundry and routes to this workflow when the prompt mentions Foundry Loom, experiment tracking, run metrics, traces, spans, OTLP ingestion, or W&B-compatible operations.

Example prompts:

Use Foundry Loom to compare the loss metric between two experiment runs.
Query detailed spans for my Foundry Loom run using this JSON filter.
Upload this OTLP protobuf trace file to my Foundry experiment run.
Send a W&B-compatible GraphQL request to my Microsoft Foundry project.

To request the workflow explicitly:

Use the microsoft-foundry skill and its experiment-tracking workflow to list my Foundry Loom runs.

The workflow verifies azure.ai.loom, requires azd >= 1.32.0, resolves the Foundry project endpoint, and selects the appropriate azd ai loom run command. Representative commands include:

azd ai loom run list
azd ai loom run compare --run-id <first-run> --run-id <second-run> --metric <metric>
azd ai loom run span query --run-id <run-id> --filter-file <filter.json>
azd ai loom run ingest traces --run-id <run-id> --file <traces.pb>

Validation

• npm run build
• cd scripts && npm run frontmatter
• cd scripts && npm run references
• cd scripts && npm run vally validate-stimulus

Dependency

The documented standalone azure.ai.loom extension and azd ai loom run command hierarchy are implemented by Azure/azure-dev#9795. Official registry publication follows that PR's release handoff.

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[Copilot]

Pull request overview

Adds an experiment-tracking sub-skill under the existing microsoft-foundry skill to guide users through inspecting Foundry experiment runs, querying spans/traces, and ingesting telemetry via azd ai project (including W&B-compatible requests), with accompanying Vally coverage.

Changes:

• Extended microsoft-foundry skill routing to include the new experiment-tracking sub-skill.
• Added new experiment-tracking workflow documentation and reference pages for run inspection, span/trace querying, and ingestion/W&B commands.
• Added new Vally stimuli covering routing and command selection for experiment tracking scenarios.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
plugins/azure-skills/skills/microsoft-foundry/SKILL.md	Registers the new experiment-tracking sub-skill and updates top-level routing description.
plugins/azure-skills/skills/microsoft-foundry/experiment-tracking/experiment-tracking.md	Adds the experiment-tracking workflow, quick reference, and error-handling guidance for azd ai project.
plugins/azure-skills/skills/microsoft-foundry/experiment-tracking/references/run-inspection.md	Documents run inspection and comparison commands.
plugins/azure-skills/skills/microsoft-foundry/experiment-tracking/references/span-trace-queries.md	Documents span query and trace-chat commands plus input-shape guidance.
plugins/azure-skills/skills/microsoft-foundry/experiment-tracking/references/ingestion-wandb.md	Documents OTLP/agent telemetry ingestion and W&B-compatible request commands.
evals/azure-skills/microsoft-foundry/eval.yaml	Adds Vally routing/response-quality stimuli for experiment tracking scenarios.

Suppressed comments (1)

plugins/azure-skills/skills/microsoft-foundry/experiment-tracking/experiment-tracking.md:36

• In the error table, the symptom mixes a full message for run with just a quoted word for ingest, which is ambiguous. Using the same error-message form for both commands will make it clearer what users should look for.

| `unknown command "run"` or `"ingest"` | Update `azd` and reinstall/update `azure.ai.projects`, then verify with `--help`. |

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[anchenyi]

The PR looks good to me.

Since you're adding a new azd extension, you'll need to merge the azd changes first and wait for the extension to be released. Once that's done, I can approve and merge this PR in Foundry Skill.

Let me know when this PR is ready to ship.

[HarshaVardhanBabu (babu-namburi)]

anchenyi
Will keep you posted once the azd changes are ready.