Workflow Planning & Cloud Operations Skills

Use when the user wants to monitor or watch an Azure DevOps (ADO) pull request build to completion — "monitor PR [url] and let me know when the build completes", "watch this PR build", "tell me when the PR build passes and a work item is linked", "is the PR gate green yet", ADO PR build/policy/gate status polling. Read-only; reports status, does not post comments or complete the PR.

Bounded, interactive engineering-readiness review of a concrete implementation plan BEFORE coding — routine features, refactors, bug fixes. Challenges scope (Step 0), then reviews architecture, code quality, tests, and performance with opinionated, recommendation-first questions. Review only — does not modify code. For an exhaustive maximum-rigor security/operations/failure-mode audit of a high-risk or cross-cutting plan use plan-mega-review, and for multi-reviewer adversarial critique of an idea, decision or artifact that is not a pre-coding plan gate use adversarial-review. Triggers: "plan exit review", "review my plan before I build", "engineering readiness review", "scope challenge", "is this plan ready to implement".

Maximum-rigor, exhaustive review of a HIGH-RISK or cross-cutting implementation plan, design doc, or architecture proposal — use only when the user explicitly asks for a deep/mega/boil-the-ocean review or the change is high-blast-radius. Three modes: SCOPE EXPANSION (build the cathedral), HOLD SCOPE (bulletproof what is here), SCOPE REDUCTION (cut to essentials). Maps every failure mode, demands observability, models threats, names error and rescue paths. Review only — does not modify code. For an ordinary bounded pre-coding review use plan-exit-review, and for multi-reviewer adversarial critique of an idea, decision or artifact that is not a pre-coding plan gate use adversarial-review. Triggers: "mega plan review", "maximum rigor review", "boil the ocean review", "review this high-risk plan", "deep review before I build this".

Turn a messy pile of sources into an inspectable "project room" and produce a grounded, source-cited deliverable. Portable and self-contained: works with no prior setup — on first use it creates a project-rooms base folder (default ~/project-rooms, override with PROJECTROOMSDIR) and your first room. Invoked with no room named, it lists existing rooms and asks which to use, with "create new project room" last. A room prepares an inspectable source inventory + duplicate/conflict/missing-context logs + summaries + working brief, STOPS for human review, THEN drafts a deliverable that cites Source IDs, labels inferences, and flags unsupported claims. It can also capture and index sources as you work. Triggers: "project room", "organize my sources", "build a source inventory", "which room is this", "index the room sources", "draft from the room", "refresh the project room", "archive superseded outputs", "new project room", "working brief". Preparation before drafting; stable IDs; never overwrites originals or invents facts.