Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Manage PR crises using classification, golden hour response, crisis statement templates (3C framework), and reputation recovery planning. Use this skill when the user faces negative media coverage, a viral complaint, product safety issues, executive misconduct, or any situation threatening brand reputation — even if they say 'we're getting destroyed on social media', 'draft a response to this article', 'how do we handle this PR disaster', or 'prepare for potential backlash'.
.claude/skills/asgard-ai-platform-pr-crisis-response/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-22 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 87% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 14% | 0% |
IRON LAW: First Response Within 1 Hour
The first hour after a crisis becomes public is the "golden hour."
Silence in this window = others control the narrative. A holding
statement ("We are aware of [issue] and are investigating. We will
provide an update by [time].") is infinitely better than silence.| Level | Description | Examples | Response | |-------|-----------|---------|----------| | Level 1: Noise | Minor complaint, limited reach | Single negative review, individual social post | Monitor, respond if needed | | Level 2: Issue | Growing attention, media interest possible | Multiple complaints on same topic, minor influencer post | Proactive response, prepare statement | | Level 3: Crisis | Widespread media coverage, significant reputation damage | Product recall, data breach, viral scandal, executive misconduct | Full crisis protocol, C-suite involvement |
Hour 1: Contain
Hours 2-6: Respond
Days 1-7: Manage
Weeks 2-4: Recover
[CONCERN] We are deeply concerned about [specific situation].
Our thoughts are with [affected parties].
[COMMITMENT] We are taking immediate action:
1. [Specific action 1]
2. [Specific action 2]
3. [Investigation/review underway]
[CONTROL] We have [crisis team/process] in place. We will provide
updates [when and where]. For questions: [contact info].| Don't | Why | Instead | |-------|-----|---------| | "No comment" | Implies guilt or indifference | "We are investigating and will share findings" | | "We're sorry IF anyone was offended" | Non-apology, dismissive | "We apologize for specific thing]" | | Blame others | Looks defensive | Take responsibility for your part | | Speculate | May be wrong, creates new problems | "Here is what we know so far" | | Minimize | Alienates affected people | Acknowledge the seriousness |
markdown# Crisis Response: {Situation} ## Classification - Level: 1 (Noise) / 2 (Issue) / 3 (Crisis) - Type: Product / Personnel / Operational / External - Current media coverage: {scope} ## Holding Statement (< 1 hour) {Draft} ## Full Statement (< 6 hours) {3C framework draft} ## Talking Points (for CS/Sales) 1. {key message} 2. {key message} 3. {redirect: "For further questions, please contact [PR]"} ## Anticipated Questions & Answers | Question | Answer | |----------|--------| | {likely question} | {prepared answer} | ## Recovery Plan 1. {corrective action + timeline}
references/social-crisis.mdreferences/reputation-recovery.md| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-22 | fail→pass | 10,382 | 13,815 | +33% | 1 | 1 | 0% | 1,674 | 3,306 | +97% | 0 | 0 | — |
case-03 | fail→pass | 20,816 | 16,663 | -20% | 1 | 1 | 0% | 3,031 | 3,539 | +17% | 0 | 0 | — |
case-01 | fail→fail | 24,981 | 34,291 | +37% | 1 | 1 | 0% | 3,688 | 3,838 | +4% | 0 | 0 | — |
case-02 | fail→pass | 27,639 | 19,758 | -29% | 1 | 1 | 0% | 3,872 | 4,022 | +4% | 0 | 0 | — |
case-04 | pass→fail | 14,111 | 19,556 | +39% | 1 | 1 | 0% | 1,972 | 4,012 | +103% | 0 | 0 | — |
case-05 | pass→pass | 11,859 | 16,150 | +36% | 1 | 1 | 0% | 1,845 | 3,576 | +94% | 0 | 0 | — |
case-06 | pass→pass | 10,422 | 10,736 | +3% | 1 | 1 | 0% | 1,723 | 2,764 | +60% | 0 | 0 | — |
case-07 | fail→pass | 11,349 | 10,917 | -4% | 1 | 1 | 0% | 1,623 | 3,028 | +87% | 0 | 0 | — |
case-08 | fail→pass | 22,515 | 17,291 | -23% | 1 | 1 | 0% | 3,049 | 3,490 | +14% | 0 | 0 | — |
case-09 | pass→pass | 10,448 | 14,290 | +37% | 1 | 1 | 0% | 1,510 | 3,164 | +110% | 0 | 0 | — |
case-10 | fail→pass | 21,053 | 14,691 | -30% | 1 | 1 | 0% | 3,121 | 3,286 | +5% | 0 | 0 | — |
case-11 | fail→pass | 7,073 | 13,084 | +85% | 1 | 1 | 0% | 1,182 | 2,953 | +150% | 0 | 0 | — |
case-12 | pass→pass | 12,886 | 14,495 | +12% | 1 | 1 | 0% | 1,693 | 3,355 | +98% | 0 | 0 | — |
case-13 | pass→pass | 12,312 | 7,139 | -42% | 1 | 1 | 0% | 1,508 | 2,238 | +48% | 0 | 0 | — |
case-14 | fail→fail | 10,744 | 13,835 | +29% | 1 | 1 | 0% | 1,627 | 3,222 | +98% | 0 | 0 | — |
case-15 | fail→fail | 16,791 | 13,965 | -17% | 1 | 1 | 0% | 2,645 | 3,233 | +22% | 0 | 0 | — |
case-16 | pass→pass | 16,682 | 19,001 | +14% | 1 | 1 | 0% | 2,216 | 3,825 | +73% | 0 | 0 | — |
case-17 | fail→pass | 14,050 | 11,677 | -17% | 1 | 1 | 0% | 2,335 | 2,951 | +26% | 0 | 0 | — |
case-18 | fail→pass | 21,275 | 15,831 | -26% | 1 | 1 | 0% | 3,117 | 3,628 | +16% | 0 | 0 | — |
case-19 | fail→fail | 14,947 | 14,370 | -4% | 1 | 1 | 0% | 2,327 | 3,107 | +34% | 0 | 0 | — |
case-20 | fail→pass | 25,751 | 16,907 | -34% | 1 | 1 | 0% | 4,057 | 3,866 | -5% | 0 | 0 | — |
case-21 | fail→fail | 20,430 | 11,759 | -42% | 1 | 1 | 0% | 3,214 | 2,919 | -9% | 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.