Show progress during long code reviews

[mwiemer-microsoft]

While running an agent session, I triggered a 20+ minute code review. I couldn't tell what the review agent was doing, so this PR adds a status log to review sessions.

[azure-pipelines]

Azure Pipelines:
2 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[mwiemer-microsoft]

@RikkiGibson requesting review

[dibarbet]

While running an agent session, I triggered a 20+ minute code review. I couldn't tell what the review agent was doing, so this PR adds a status log to review sessions.

FWIW in the github copilot app I'm able to click to open the review sessions and see the live progress:

so for me this would just be using extra tokens. Not strictly opposed though

[Copilot]

🟢 Approval recommended

The only finding is a non-blocking nit about updates during uninterruptible commands.

Pull request overview

Adds progress-status guidance to long-running code review sessions.

Changes:

• Adds start, phase-transition, timed, and delegated-work updates.
• Keeps progress messages separate from final findings.

File summaries

File	Summary	Review finding
.github/skills/code-review/SKILL.md	Documents review-session progress reporting requirements.	Nit (1 vote): The five-minute guarantee may not hold during uninterruptible commands; consider heartbeats or bounded operations.

Review details

Suppressed comments (1)

.github/skills/code-review/SKILL.md:29

• This five-minute guarantee cannot be met while the agent is blocked inside one long-running tool/command: the model cannot emit another session message until that call returns, so a 20+ minute review can still be silent despite this instruction. Please either add a session/runner heartbeat that can stream updates during waits, or change the rule to require bounded/chunked operations with updates before and after each operation rather than promising periodic updates during an uninterruptible wait.

3. If any phase or delegated work runs longer than five minutes, print a concise progress update at least every five minutes. Say what is complete, what is still running, and whether the review is blocked or proceeding.
4. Before waiting on delegated agents or long-running commands, say what is being awaited and why. When they finish, report the result before continuing.

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

[Copilot]

🟡 Changes recommended

The progress cadence is unbounded, and the repo-wide convention is not documented in the required memory conventions file.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review details

Suppressed comments (1)

.github/skills/code-review/SKILL.md:26

• This establishes a repo-wide agent behavior, but the repository’s documentation policy requires new repo-wide patterns to be added to .github/memory/CONVENTIONS.md (.github/copilot-instructions.md:97, .github/skills/update-agent-docs/SKILL.md:21). Without that entry, agents that load the memory conventions but do not select this skill will miss the progress-update requirement; please add the convention there or explain why this skill is intentionally the sole source.

### Review Progress Updates

1. Print your plan before beginning
2. After each step of your plan, print a quick result and your next step
3. Print something at least every few minutes so humans can audit activity

• Files reviewed: 1/1 changed files
• Comments generated: 1
• Review effort level: Lite