Credential Sprawl

Front API token was recovered from a backup instead of the live config. Pattern: Teamwork token recovery (March 18), GSC reauth (multiple times), Slack token search. Credentials are scattered across backups, project files, and manual notes with no unified discovery path.

This isn't a technical failure — the credentials exist, recovery works. But it's a knowledge problem: sessions don't have easy centralized access. Results: time waste during execution (hunt the backup), team deployment friction (coworkers don't know where tokens are), and rotation risk (manual refresh, easy to miss a service).

Improvement #19 (Google token validator) is a step toward proactive refresh. But credential discovery is the real pain. For team scale: unified vault or documented locations, refresh patterns codified per service, PERMISSIONS.md updated to include recovery steps for every tool.

Mission Control v0.8 shipped five major feature batches in six hours — sort, Teamwork integration, monitoring, page split, export script. But deployment is blocked on: OAuth not implemented, worker not in prod account, KV secrets not migrated, CORS wide open. Features are polished. Deployment is rough.

The gap explains why "deployment timing" is always ambiguous. When is something "ready to deploy"? The current answer is implicit: when features feel complete. The better answer: when ops are ready. Those are different gates, and only one is currently tracked.

Making "Deployment Readiness" an explicit tracked item — separate from feature work, with its own checklist (OAuth, CORS, backups, alerting, secrets rotation) — removes the ambiguity and gives the team a clear go-live signal.

improvements.md is at 29 entries and 37KB. Entries range from 🟢 confirmed (Improvements #6, #8, #11) to 💡 design (Improvements #24–29). But there's no signal on which improvements are still relevant, which should be archived, and whether evaluation time is being spent on live ideas or stale ones.

Improvement #27 (measurement) will address "which improvements work?" But there's a downstream problem it doesn't solve: "which improvements are obsolete?" Without archival, the active queue accumulates indefinitely — eventually causing decision fatigue and hiding the most important items in noise.

Solution: quarterly "improvement hygiene" step that audits the active queue, identifies zero-reference improvements from the past 60 days, evaluates relevance, and archives obsolete ones to improvements-archive.md. Keep the active queue lean.