Add durable session-to-work-log capture for delivery and PR hardening

[phaedrus-brandt]

Problem

The Apollo a11y session required manual work-log updates after major branch points: MCP recovery, PR openings, hardening fixes, CI outcomes, mergeability follow-up, and residual blockers. The harness did not provide a first-class way to continuously capture these events while work progressed.

Observed gaps

• Work-log notes were manually composed after the fact.
• Important state could have been lost if the session ended before final synthesis.
• Apollo updates required ad hoc API scripting when MCP mutation risked overwriting long descriptions.
• Final state needed multiple systems reconciled: local verification, GitHub PR status, Azure checks, subagent findings, and Apollo tickets.

Desired workflow enhancement

Add a delivery journal/work-log capture mechanism that can be invoked throughout /deliver, /qa, /code-review, /refactor, /mergeability-preflight, and /pr-polish.

It should capture:

• ticket IDs and PR URLs
• branch/worktree/head SHA
• commands run and results
• subagent findings and fixes
• CI/Azure status transitions
• decisions and residual risks
• no-merge/no-deploy boundary confirmations

Acceptance criteria

• Can append to Apollo or a local handoff file without overwriting existing long descriptions.
• Uses structured sections and timestamps.
• Has a dry-run mode for review before mutation.
• Supports multiple tickets/PRs in one session.
• Emits a final handoff summary suitable for next-session recovery.