raid-facilitation

Skill: raid-facilitation

Purpose: Turn a RAID stub into a register that actually drives mitigation spend to where the expected loss is. Goes deeper than ravenclaude-core/project-manager's lightweight RAID-log hygiene. Used by risk-and-raid-analyst.

When to use

• Standing up a risk register for a new project.
• Refreshing RAID after a scope change, a missed milestone, or a new dependency.
• Quantifying a high-stakes risk (go/no-go-moving).
• Triaging a backlog of issues to owners + dates.

The procedure

Risks

1. Frame cause → event → consequence. "Server might fail" is not a risk. "Because the gateway is single-node (cause), it may go down during refresh (event), delaying month-end close (consequence)" is.
2. Score against a stated rubric. Qualitative probability × impact with the scale written down (don't let "High" float undefined). Aggregate related low risks over the same objective — several minors can combine into a major.
3. Quantify where the stakes justify it. For high-impact risks: an impact range + probability → expected value (EMV) → the contingency reserve it justifies. Don't quantify trivia; do quantify the ones that move the decision.
4. A response per risk: mitigate / transfer / avoid / accept — each with a single owner and (for accept) an explicit, documented decision. A scored risk with no response is half an entry.
5. A trigger / review date per risk, so it's re-checked, not filed and forgotten.

Issues, Assumptions, Dependencies

• Issues (already occurred) — triage by severity × urgency, assign owner + due date, escalate the ones breaching threshold with a recommended path.
• Assumptions — log them; an unstated assumption is a silent risk.
• Dependencies — especially cross-team ones (the most common silent schedule risk). Name the owner on both sides + the needed-by date.

Anti-patterns this skill prevents

• Risks as bare nouns; "High" with no rubric; a scored risk with no response or owner.
• Quantifying trivia while the go/no-go risk sits as a bare "Medium."
• Contingency reserve set by gut, not by the quantified risk it covers.
• Issues logged but never triaged; cross-team dependencies absent from the register.

Output

A risk register + issue log (see ../../templates/risk-register.md): cause-event-consequence risks, the scoring rubric, responses + owners + review dates, triaged issues, and tracked assumptions/dependencies. End with the risk-and-raid-analyst Output Contract block; route contingency into the baseline via delivery-lead.