Python: Integrate message injection into harness agent

[westey-m]

Motivation & Context

The .NET HarnessAgent enables message injection by default (via
UseMessageInjection), letting a user queue additional input while an agent
turn is already running so it is picked up within the same run. The Python
harness lacked this. This PR brings the Python create_harness_agent and the
harness console sample to parity, so users can nudge/steer a running harness
agent without waiting for the current turn to finish.

Description & Review Guide

• What are the major changes?

  • create_harness_agent now always wires in MessageInjectionMiddleware
    (mirroring the .NET harness's UseMessageInjection). It is always on with
    no opt-out because it is a no-op when no messages are queued for the
    session.
  • The harness console sample (agent_runner.py, app.py, state_driver.py,
    components/agent_status.py) now supports injecting input during a
    streaming run: text submitted mid-turn is enqueued via the middleware and
    drained into the ongoing run; the queued-message display is kept in sync and
    consumed messages are echoed to the output area.
  • The planning-output observer was made robust to multi-response runs (message
    injection can trigger a follow-up model invocation): only the last
    response's text is retained for structured-output (JSON) parsing, earlier
    responses are flushed as plain text. Boundaries are detected via non-None
response_id changes, matching Python's streaming semantics.
  • Test and doc-sample touch-ups accompanying the wiring.

• What is the impact of these changes?

  • Message injection is available by default for harness agents, with no API
    surface change required from callers.
  • The behavior is additive and a no-op when nothing is queued, so existing
    harness runs are unaffected.

• What do you want reviewers to focus on?

  • The always-on/no-opt-out decision for MessageInjectionMiddleware and its
    parity with the .NET UseMessageInjection default.

Related Issue

Fixes #7026

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) — a workflow keeps the label and title prefix in sync automatically.

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework/_harness
_agent.py	118	4	96%	191, 557–558, 560
TOTAL	44099	5270	88%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
8834	33 💤	0 ❌	0 🔥	2m 16s ⏱️

[Copilot]

Pull request overview

Integrates always-on message injection into the Python harness agent and updates the harness console sample to support queueing user input mid-stream, bringing behavior closer to the .NET harness.

Changes:

• Wire MessageInjectionMiddleware into create_harness_agent by default and document the behavior.
• Update the harness console runner/UI to enqueue streaming-time input and display/echo queued messages.
• Harden planning-mode structured-output collection for multi-response runs by tracking response_id boundaries; add/adjust tests and sample touch-ups.

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
python/packages/core/agent_framework/_harness/_agent.py	Always appends MessageInjectionMiddleware to harness agent middleware stack.
python/packages/core/AGENTS.md	Documents default harness wiring including always-on message injection.
python/packages/core/tests/core/test_harness_agent.py	Adds/updates tests asserting message injection is wired by default and middleware ordering expectations.
python/samples/02-agents/harness/console/agent_runner.py	Adds streaming-time input enqueueing and queued-message syncing/echoing during streaming.
python/samples/02-agents/harness/console/app.py	Routes text submissions during streaming to the runner and updates status/placeholder behavior.
python/samples/02-agents/harness/console/components/agent_status.py	Adds queued-message status line rendering.
python/samples/02-agents/harness/console/observers/base.py	Updates observer typing/docs to use AgentResponseUpdate metadata.
python/samples/02-agents/harness/console/observers/planning_output.py	Switches to on_response_update and tracks response_id boundaries to keep only the last response for JSON parsing.
python/samples/02-agents/harness/console/state_driver.py	Extends UX protocol + simple console driver with queued-message display hook.
python/samples/02-agents/security/email_security_example.py	Adjusts sample string quoting in scenario output.
python/samples/02-agents/harness/build_your_own_claw/claw_step01_meet_your_claw.py	Minor whitespace/formatting touch-up.
python/samples/02-agents/harness/build_your_own_claw/claw_step02_working_with_data.py	Minor whitespace/formatting touch-up.
python/samples/02-agents/harness/build_your_own_claw/claw_step03_scaling_capabilities.py	Minor import ordering/whitespace touch-up.

[github-actions]

Automated Code Review

Reviewers: 5 | Confidence: 86% | Result: All clear

Reviewed: Correctness, Security Reliability, Test Coverage, Failure Modes, Design Approach

---

Automated review by westey-m's agents