Python: Retry the Magentic progress ledger when it names an unknown agent

[Lanre Shittu (Shizoqua)]

Motivation & Context

StandardMagenticManager.create_progress_ledger says it avoids selecting a non-existent agent, but the next_speaker in the ledger was never checked. When the model named a participant slightly wrong, such as "Coder" for "coder", the orchestrator treated it as invalid and went straight to the final answer, ending the run with the task unfinished.

Description & Review Guide

• What are the major changes? Inside the existing retry loop, a ledger whose next_speaker is not a participant now raises, so it is retried like a parse failure. The check is skipped when the request is already satisfied.
• What is the impact of these changes? A small naming slip from the model gets another attempt instead of ending the run early. Ledgers with valid names are unaffected.
• What do you want reviewers to focus on? If every retry still names an unknown agent, the existing failure path applies and the orchestrator resets and replans.

Related Issue

Fixes #9178

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 a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix), and a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Eduard van Valkenburg (eavanvalkenburg)]

/review

[github-actions]

MAF Automated Review — Iteration 1

Result: Findings reported
Scope: full PR (1 commit(s)): 45c45d921517
Model: gpt-5.6-sol

Overview

The change validates unsatisfied progress ledgers against configured participant names and reuses the existing bounded per-call parse retry mechanism. It preserves satisfied-ledger handling, empty-team validation, and configured round/reset limits, with a focused test for recovery on the second response. However, persistent invalid output now enters the orchestrator's reset/replan path, whose limits are unlimited by default, creating an unbounded paid-call loop.

Reviewed the supplied pull-request change set across correctness, security/reliability, architecture, and failure behavior.
1 verified finding remained after source verification (1 high) across 1 file. Details are attached to the affected lines below.

Affected areas: python/packages/orchestrations/agent_framework_orchestrations/_magentic.py