magnus919_agent-skills/technical-program-management/references/assurance-and-intervention.md

Program assurance and intervention

Applicability: Load when a program is high-consequence, materially off-track, disputed, or approaching a major tranche, transition, or funding decision.

Assurance is independent of delivery

Assurance asks whether the program's evidence, governance, forecasts, dependencies, and benefits case are credible. It is not a status meeting and should not be performed only by the team seeking approval. Define the assurance scope, reviewer independence, evidence requested, decision to inform, and response owner. Independence is a control boundary, not a claim that an external review is automatically correct.

At a proportionate checkpoint, inspect:

• mandate, authority, funding, and continued justification;
• component scope, integrated architecture/dependencies, and shared capacity;
• actual progress and forecast logic, including contradictory status;
• technical, security, safety, compliance, supplier, and operational readiness evidence;
• benefit dependency map, adoption signals, disbenefits, and measurement ownership;
• open risks/issues, decision latency, escalation history, and residual obligations.

Record findings as evidence, interpretation, limitation, recommendation, owner, and due date. The program sponsor decides whether to accept risk or change the mandate; an assurance reviewer does not silently become the decision-maker.

Troubled-program intervention

Trigger a bounded intervention when integrated outcomes breach tolerance, critical dependencies lack credible recovery, benefits assumptions fail, authority is absent, or repeated reports conflict with observable evidence. Freeze neither all work nor the historical record by default.

1. Name the intervention sponsor and decision horizon.
2. Preserve the original mandate, baselines, commitments, and variance.
3. Reconstruct components, dependencies, capacity, funding, benefits, and evidence gaps.
4. Identify the limiting constraint and distinguish symptoms from causes.
5. Present recovery, reshape, pause, transfer, or stop options with consequences.
6. Obtain an authorized decision and record residual risks and conditions.
7. Recheck at a dated or event-triggered evidence checkpoint. Rebaseline only when authorized.

If three diagnostic passes produce no new evidence, stop and escalate the missing input and affected decision. Do not convert assurance into ceremony or use a new dashboard to disguise an unchanged constraint.

Boundaries

Route detailed project recovery to technical-project-management, release evidence to production-readiness, live incidents to site-reliability-engineering, and organizational design to org-design. This skill coordinates program-level intervention and assurance inputs across them.