Assemble everything the change advisory board needs into one skimmable pack: what's
waiting for a decision, what's riskiest, what collides, and how last week's approvals
actually turned out. The outcomes section is what keeps a CAB honest.
1. Establish the review window (default: changes pending decision now, plus changes
executed since the last CAB — ask for the desk's cadence if unknown).
2. PENDING CHANGES: search the change board for approval-pending statuses.
For each, pull the one-line what/scope/window, the risk classification and scores
(from change-risk-assessment — if a change reaches the pack unclassified, list it
under "not ready: unassessed" rather than guessing a rank), and the prerequisites-gate
status (incomplete requests are listed as "not ready" with the gap, not queued for
decision).
3. RISK RANKING: order the ready changes by risk — HIGH blast radius first, then LOW
rollback confidence as tiebreaker. State the ordering rule in the pack.
4. COLLISION FLAGS: cross-check proposed windows against each other and against
already-scheduled changes (same systems, same clients, shared infrastructure, freeze
windows — reuse the checks from change-calendar-management). Flag each collision inline
on both changes involved.
5. LAST WEEK'S OUTCOMES: for changes executed since the last CAB, report per change:
implemented as approved / implemented with deviation / rolled back / not attempted —
based on ticket evidence; a bare "done" is UNVERIFIED, not success. Include the
emergency changes that bypassed the board (from emergency-change-handling) with their
retro-documentation status.
6. Compose the pack, ~1 page, in fixed order: (a) decision queue with risk rank and
one-line asks, (b) not-ready list with reasons, (c) collision flags, (d) last week's
outcomes with an honest success/rollback/unverified count, (e) recurring themes worth
a policy tweak.
7. Deliver in chat for the organizer; on request post a plain-text copy as a note on the
CAB/agenda ticket.
Guardrails: the pack informs decisions; it makes none — no change is marked approved
because it appeared in a brief. Never promote an unassessed or prerequisites-incomplete
change into the decision queue to make the pack look fuller — "not ready" is a legitimate
section. Outcome reporting uses evidence standards, not optimism: unverified completions
are reported as unverified, rollbacks are never euphemized into "partial success."
Aggregate honestly under result caps — if a board search may be capped, state the pack
may be incomplete. Sanitized for the room: name systems and clients only as the desk's
tickets do; no credentials or personal data ride along.
Magic Library