Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build a crisis communications plan to respond fast and credibly when something goes wrong. Use when asked to handle a PR crisis, draft a crisis comms plan, respond to a public backlash/scandal/incident, or prepare holding statements. Produces a crisis comms plan — situation assessment, stakeholder map, a message house, channel-by-channel statements, a holding statement, an internal brief, and a follow-up timeline.
.claude/skills/mohitagw15856-pr-crisis-response/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 140% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 283% | 0% |
In a crisis, silence reads as guilt and a clumsy statement makes it worse. The first hour decides the narrative. This skill produces a coordinated response — what you say, to whom, on which channel, and in what order — anchored in one consistent message so the company speaks with a single voice while the facts are still moving.
You'll often get the situation in a sentence ("a customer's data was exposed and it's trending"). Produce the full plan anyway — infer the likely stakeholders, channels, and questions, label assumptions, and clearly flag where facts must be confirmed before publishing. Never stall for complete information; a crisis plan with labelled unknowns beats no plan. Mark anything legally sensitive for review.
Ask for these only if they aren't already provided (else infer and label):
1. Situation assessment — the facts (confirmed / unconfirmed / unknown), severity, and the likely trajectory.
2. Guiding principles — be fast, honest, human, and consistent; lead with the people affected, not the company.
3. Stakeholder map — who needs to hear from you, in priority order, and what each one needs:
| Audience | What they care about | Channel | Priority | |---|---|---|---| | Affected customers | am I harmed, what now | direct email / in-app | 1 | | Employees | what do I tell people | internal note | 1 | | Press / public | what happened, accountability | statement / social | 2 | | Regulators / partners | obligations, next steps | direct, formal | as required |
4. Message house — the single core message (one sentence), three supporting pillars (accountability, action, prevention), and the facts that back each. Everything else stays consistent with this.
5. Holding statement — a short, publishable-now statement that acknowledges, shows you're acting, and commits to an update by a stated time — without speculating or admitting unverified fault.
6. Channel statements — tailored versions for the priority channels (customer email, social post, press statement, internal brief), each on-message.
7. Q&A prep — the hardest questions you'll be asked and honest, on-message answers (incl. "what we don't yet know").
8. Follow-up timeline — when the next update comes, who owns it, and the criteria for standing down.
Crisis communications practice — single-source-of-truth messaging, stakeholder prioritisation, holding statements, and accountable, people-first response.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 49,047 | 34,228 | -30% | 1 | 1 | 0% | 6,824 | 5,900 | -14% | 0 | 0 | — |
case-02 | fail→pass | 45,046 | 28,406 | -37% | 1 | 1 | 0% | 7,181 | 5,716 | -20% | 0 | 0 | — |
case-03 | fail→fail | 41,806 | 34,695 | -17% | 1 | 1 | 0% | 5,627 | 5,873 | +4% | 0 | 0 | — |
case-04 | fail→pass | 22,854 | 27,353 | +20% | 1 | 1 | 0% | 2,562 | 4,471 | +75% | 0 | 0 | — |
case-05 | pass→pass | 26,524 | 21,119 | -20% | 1 | 1 | 0% | 3,009 | 4,063 | +35% | 0 | 0 | — |
case-06 | fail→pass | 23,964 | 36,964 | +54% | 1 | 1 | 0% | 2,503 | 6,009 | +140% | 0 | 0 | — |
case-07 | pass→pass | 23,354 | 31,688 | +36% | 1 | 1 | 0% | 2,767 | 5,052 | +83% | 0 | 0 | — |
case-08 | fail→pass | 16,792 | 30,128 | +79% | 1 | 1 | 0% | 2,535 | 4,381 | +73% | 0 | 0 | — |
case-09 | fail→pass | 12,613 | 26,335 | +109% | 1 | 1 | 0% | 1,128 | 4,325 | +283% | 0 | 0 | — |
case-10 | pass→pass | 23,904 | 32,092 | +34% | 1 | 1 | 0% | 2,415 | 4,540 | +88% | 0 | 0 | — |
case-11 | fail→fail | 21,325 | 36,933 | +73% | 1 | 1 | 0% | 2,498 | 5,699 | +128% | 0 | 0 | — |
case-12 | fail→pass | 21,146 | 55,751 | +164% | 1 | 1 | 0% | 2,079 | 5,351 | +157% | 0 | 0 | — |
case-13 | fail→pass | 14,829 | 17,350 | +17% | 1 | 1 | 0% | 2,037 | 2,619 | +29% | 0 | 0 | — |
case-14 | fail→fail | 25,080 | 33,476 | +33% | 1 | 1 | 0% | 2,692 | 5,157 | +92% | 0 | 0 | — |
case-15 | fail→pass | 17,871 | 29,266 | +64% | 1 | 1 | 0% | 1,889 | 4,360 | +131% | 0 | 0 | — |
case-16 | fail→pass | 11,675 | 28,897 | +148% | 1 | 1 | 0% | 1,549 | 4,152 | +168% | 0 | 0 | — |
case-17 | pass→pass | 25,235 | 38,977 | +54% | 1 | 1 | 0% | 2,770 | 5,557 | +101% | 0 | 0 | — |
case-18 | fail→pass | 19,752 | 30,178 | +53% | 1 | 1 | 0% | 2,233 | 4,790 | +115% | 0 | 0 | — |
case-19 | pass→pass | 19,144 | 31,801 | +66% | 1 | 1 | 0% | 2,269 | 4,402 | +94% | 0 | 0 | — |
case-20 | pass→pass | 13,781 | 26,502 | +92% | 1 | 1 | 0% | 1,472 | 4,050 | +175% | 0 | 0 | — |
case-21 | pass→fail | 24,855 | 36,999 | +49% | 1 | 1 | 0% | 3,418 | 5,740 | +68% | 0 | 0 | — |
case-22 | pass→pass | 24,737 | 30,310 | +23% | 1 | 1 | 0% | 2,896 | 4,210 | +45% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.