Python: add Magentic custom manager prompts sample

[NekoPunch (orangeCatDeveloper)]

Motivation & Context

MagenticBuilder accepts an override for each of StandardMagenticManager's seven built-in prompts, but no sample shows it, so users with a required report format have to read _magentic.py to find the knobs and the placeholders available to each prompt.

Description & Review Guide

• What are the major changes?

  A new sample, python/samples/03-workflows/orchestrations/magentic_custom_prompts.py, runs an incident review board and overrides six of the seven prompts. Two design points:

  The replan prompts are overridden alongside the initial ones because a stall makes the manager rebuild both the fact sheet and the plan — overriding only the initial prompts would silently drop the custom format mid-run.

  progress_ledger_prompt is deliberately left at its default: its response is parsed as JSON, so an override must reproduce the built-in schema. The docstring says so, along with a table of every overridable prompt and its available placeholders. Overrides may omit placeholders they do not need.

• What is the impact of these changes?

  Samples only, plus one README row.

• What do you want reviewers to focus on?

  Whether the prompt/placeholder table still matches the StandardMagenticManager contract.

Ruff and targeted Pyright checks are clean.

To exercise the same path a sample user follows, I created a temporary Foundry V2 project in East US 2, deployed gpt-4.1-mini, and authenticated with AzureCliCredential. I then ran:

FOUNDRY_PROJECT_ENDPOINT=<temporary-project-endpoint> \
FOUNDRY_MODEL=<temporary-deployment-name> \
uv run python samples/03-workflows/orchestrations/magentic_custom_prompts.py

The process exited successfully after emitting PLAN_CREATED with the custom INCIDENT REVIEW BRIEF, streaming participant responses, and producing a final report with the requested SUMMARY, FINDINGS, and RECOMMENDATION sections. The temporary deployment, project, and resource group were deleted afterward.

The nightly validate-03-workflows job is currently disabled, so this was a manual end-to-end run.

Related Issue

Fixes #7298

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is breaking, add the breaking change label or put [BREAKING] in the title prefix — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Pull request overview

Adds a Python sample demonstrating custom Magentic manager prompts for an incident-review workflow.

Changes:

• Overrides six planning, replanning, ledger, and final-answer prompts.
• Streams participant and task-ledger output.
• Adds the sample to the orchestration README.

Reviewed changes

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

File	Description
python/samples/03-workflows/orchestrations/README.md	Links the new sample.
python/samples/03-workflows/orchestrations/magentic_custom_prompts.py	Implements and documents custom manager prompts.

---

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