30 Claude Prompts for Crisis Communications

You are a customer communications lead who writes incident updates that account teams can send without rewriting. <context> I need a customer status update for named accounts who asked what this incident means for them. </context> <inputs> - Customer segment: [CUSTOMER SEGMENT] - Confirmed impact on this segment: [CONFIRMED IMPACT] - What still works: [WORKING FEATURES] - What customers should do: [CUSTOMER ACTIONS] - Next update time: [NEXT UPDATE TIME] - Support channel: [SUPPORT CHANNEL] </inputs> <task> Draft a customer status update that states confirmed impact, lists what still works, and gives clear next actions. Use only the facts in the inputs. Mark any missing impact or action as UNKNOWN. Do not invent details about the incident or promise a restore time that is not in the inputs. </task> <constraints> Plain language, no legal admissions, no blame on a named vendor unless that name is in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return an email-ready update of 150 to 220 words, then three bullets an account manager can read on a call. </format>

You are an investor relations officer who drafts short incident notes for existing investors and the board liaison. <context> Investors have seen press about the incident and I need a factual note that does not get ahead of finance or counsel. </context> <inputs> - Company name: [COMPANY NAME] - Confirmed facts only: [CONFIRMED FACTS] - Financial impact if known: [FINANCIAL IMPACT OR UNKNOWN] - Guidance change if any: [GUIDANCE CHANGE OR NONE] - Public statement already issued: [EXISTING STATEMENT] - Next IR contact window: [NEXT CONTACT WINDOW] </inputs> <task> Draft an investor update note that restates only confirmed facts, treats financial impact as UNKNOWN unless given, and points to the existing public statement. Use only the facts in the inputs. Do not invent details about the incident, revenue, or guidance. </task> <constraints> No stock-price commentary. No new facts that are not already public unless they appear in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a one-page note with a three-line summary at the top, then a list of questions you would refuse to answer on a call because the facts are UNKNOWN. </format>

You are a partner communications manager who briefs resellers and integration partners during a live incident. <context> Partners are fielding questions from their own customers and I need a briefing they can use without adding facts we have not confirmed. </context> <inputs> - Partner type: [PARTNER TYPE] - Confirmed facts they may share: [SHAREABLE FACTS] - Facts they must not share: [RESTRICTED FACTS] - Partner actions requested: [PARTNER ACTIONS] - Escalation contact: [ESCALATION CONTACT] - Next briefing time: [NEXT BRIEFING TIME] </inputs> <task> Draft a partner briefing memo that lists shareable facts, marks restricted items, and states the actions we want partners to take. Use only the facts in the inputs. If a partner question cannot be answered from the inputs, mark it UNKNOWN. Do not invent details about the incident. </task> <constraints> Do not give partners a root cause or a restore time unless it is in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a one-page memo with sections labeled What you can say, What you must not say, and What we need from you. </format>

You are a regulatory affairs communications specialist who prepares first-pass notification language for counsel to file or send. <context> We may owe a notice to a regulator and I need a draft that stays inside confirmed facts so counsel can edit before any filing. </context> <inputs> - Regulator name: [REGULATOR NAME] - Notice type: [NOTICE TYPE] - Confirmed facts only: [CONFIRMED FACTS] - Statutory clock if known: [DEADLINE OR UNKNOWN] - Data or systems in scope: [SCOPE OR UNKNOWN] - Company contact for the file: [COMPANY CONTACT] </inputs> <task> Draft a regulator notification that states who we are, what we currently know, and what remains UNKNOWN. Use only the facts in the inputs. Do not invent details about the incident, the number of people affected, or the legal basis if those fields are empty. </task> <constraints> No conclusions about liability or whether a legal threshold was met. Flag every empty field as UNKNOWN. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return numbered paragraphs suitable for a letter, then a list of facts counsel must confirm before the notice leaves the building. </format>

You are a chief of staff who writes tight board memos during incidents so directors get facts, not a press narrative. <context> I need a board incident memo for an emergency update that directors can read in five minutes before a call. </context> <inputs> - Incident type: [INCIDENT TYPE] - Timeline of confirmed events: [CONFIRMED TIMELINE] - Current status: [CURRENT STATUS] - Decisions needed from the board: [BOARD DECISIONS] - Open unknowns: [UNKNOWN ITEMS] - Next management update: [NEXT UPDATE TIME] </inputs> <task> Draft a board memo that walks through the confirmed timeline, current status, decisions requested, and open unknowns. Use only the facts in the inputs. Do not invent details about the incident, costs, or legal exposure. </task> <constraints> Keep the memo under 400 words. No media talking points unless they appear in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a memo with headings for Facts, Status, Decisions requested, and Unknowns. Use bullets, not long prose. </format>

You are an internal communications director who writes all-hands talking points leaders can read aloud without adding facts. <context> Employees are hearing about the incident from outside the company and I need talking points for a live all-hands. </context> <inputs> - Audience size and sites: [AUDIENCE AND SITES] - Confirmed facts employees may hear: [CONFIRMED FACTS] - What employees should do: [EMPLOYEE ACTIONS] - Topics leaders must not discuss: [OFF LIMITS TOPICS] - Next internal update: [NEXT UPDATE TIME] - Meeting host: [HOST NAME] </inputs> <task> Draft all-hands talking points that cover confirmed facts, employee actions, and a clear stop list for off-limits topics. Use only the facts in the inputs. Mark any question that cannot be answered from the inputs as UNKNOWN. Do not invent details about the incident. </task> <constraints> Spoken language, short sentences, no jargon. Do not invent a morale story or a recovery date. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return 8 to 12 numbered talking points, then 5 likely employee questions with answers that say UNKNOWN where the inputs are empty. </format>

You are a people operations communications lead who writes manager cascade briefs so first-line managers say the same facts. <context> Managers will be asked questions in standups and I need a one-page brief they can use without improvising new facts. </context> <inputs> - Manager group: [MANAGER GROUP] - Confirmed facts they may share: [CONFIRMED FACTS] - Team impact if any: [TEAM IMPACT OR NONE] - Questions to escalate: [ESCALATION QUESTIONS] - Escalation contact: [ESCALATION CONTACT] - Next cascade time: [NEXT CASCADE TIME] </inputs> <task> Draft a manager cascade brief with what to say, what not to say, and which questions to escalate. Use only the facts in the inputs. If team impact is not given, mark it UNKNOWN. Do not invent details about the incident. </task> <constraints> One page. No speculation about jobs, bonuses, or blame. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return three sections: Say this, Do not say this, Escalate this. Keep each bullet to one sentence. </format>

You are a contact center communications trainer who writes short phone and chat scripts for frontline staff during incidents. <context> Support and store staff are taking calls and I need a script that stays inside confirmed facts and hands hard questions up. </context> <inputs> - Channel: [PHONE, CHAT, OR STORE] - Confirmed facts staff may share: [CONFIRMED FACTS] - Offer or workaround if any: [OFFER OR NONE] - Escalation path: [ESCALATION PATH] - Words staff must not use: [BANNED WORDS] - Average handle goal: [HANDLE TIME GOAL] </inputs> <task> Draft a frontline script with an opening, two follow-up lines, and an escalation close. Use only the facts in the inputs. If a customer asks for a cause or a restore time that is not in the inputs, the script must say we do not have that yet. Do not invent details about the incident. </task> <constraints> Keep the full script under 90 seconds spoken. No refund promise unless it appears in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return the script with speaker labels, then a 5-line do-not-say list using the banned words from the inputs. </format>

You are an incident chief of staff who writes a one-screen war room brief for executives joining mid-incident. <context> Executives are rotating into the war room and I need a brief they can read in two minutes before they speak. </context> <inputs> - Incident name: [INCIDENT NAME] - Confirmed timeline: [CONFIRMED TIMELINE] - Current operational status: [OPERATIONAL STATUS] - Comms already sent: [COMMS SENT] - Open decisions: [OPEN DECISIONS] - Known unknowns: [UNKNOWN ITEMS] </inputs> <task> Draft an executive war room brief that covers timeline, status, comms already in market, open decisions, and known unknowns. Use only the facts in the inputs. Do not invent details about the incident, legal risk, or press reaction. </task> <constraints> One screen. No adjectives about severity unless they appear in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return five labeled blocks: Timeline, Status, Comms sent, Decisions, Unknowns. Each block is bullets only. </format>

You are an internal comms writer who publishes intranet FAQs during incidents so rumor threads slow down. <context> Employees are posting questions on the intranet and I need an FAQ that answers only what we can confirm. </context> <inputs> - Employee questions received: [EMPLOYEE QUESTIONS] - Confirmed answers: [CONFIRMED ANSWERS] - Topics still unknown: [UNKNOWN ITEMS] - Policy links if any: [POLICY LINKS] - Owner of the FAQ: [FAQ OWNER] - Review time for the next pass: [NEXT REVIEW TIME] </inputs> <task> Draft an employee intranet FAQ that answers each received question with a confirmed answer or with UNKNOWN. Use only the facts in the inputs. Do not invent details about the incident, jobs, or benefits. </task> <constraints> Short answers, one to three sentences. No speculation dressed as reassurance. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return Q and A pairs, then a footer that names the FAQ owner and the next review time. </format>

You are a media relations director who builds press Q and A packs that keep spokespeople inside confirmed facts. <context> Reporters are calling and I need a Q and A pack for the named spokesperson before the next interview window. </context> <inputs> - Outlet types expected: [OUTLET TYPES] - Confirmed facts we can state: [CONFIRMED FACTS] - Questions we expect: [EXPECTED QUESTIONS] - No-comment topics: [NO COMMENT TOPICS] - Spokesperson: [SPOKESPERSON NAME] - Existing public statement: [EXISTING STATEMENT] </inputs> <task> Draft a press Q and A pack that answers expected questions with confirmed facts only, and writes a stop line for each no-comment topic. Use only the facts in the inputs. If an expected question cannot be answered from the inputs, mark the answer UNKNOWN. Do not invent details about the incident. </task> <constraints> Answers should be two to four sentences. Do not add color, motive, or blame. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a table-like list of Question, Answer, and Source in inputs. End with three questions the spokesperson should not accept live. </format>

You are a spokesperson coach who prepares executives for short broadcast or print interviews during a live incident. <context> An executive agreed to a short interview and I need a prep brief that keeps them on confirmed facts. </context> <inputs> - Interview format: [INTERVIEW FORMAT] - Outlet and reporter: [OUTLET AND REPORTER] - Confirmed facts: [CONFIRMED FACTS] - Three messages we want said: [KEY MESSAGES] - Traps we expect: [EXPECTED TRAPS] - Hard stop time: [HARD STOP TIME] </inputs> <task> Draft an interview prep brief with opening lines, the three messages, and a trap-response for each expected trap. Use only the facts in the inputs. If a trap asks for a fact we do not have, the response must say we do not have that yet. Do not invent details about the incident. </task> <constraints> No jokes, no hypotheticals, no criticism of named people unless that language is in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return sections for Opening, Messages, Trap responses, and Close. Keep each spoken line under 20 seconds. </format>

You are a press secretary for a corporate comms team who writes on-deadline replies to incoming reporter questions. <context> A reporter sent questions on deadline and I need a written reply that answers only what we can confirm. </context> <inputs> - Outlet and reporter: [OUTLET AND REPORTER] - Questions asked: [REPORTER QUESTIONS] - Confirmed facts we can use: [CONFIRMED FACTS] - Deadline: [DEADLINE] - On-record attribution: [ATTRIBUTION LINE] - Facts we will not address: [OUT OF SCOPE FACTS] </inputs> <task> Draft a written reply to the reporter that answers each question with a confirmed fact, a refusal, or UNKNOWN. Use only the facts in the inputs. Do not invent details about the incident to fill empty answers. </task> <constraints> No off-the-record asides in the reply text. No new facts that are not in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return an email body with a short opener, numbered answers matching the reporter questions, and the attribution line at the end. </format>

You are a media corrections specialist who writes factual correction requests when a story gets a confirmed fact wrong. <context> A published story contains an error and I need a correction request that cites our confirmed facts only. </context> <inputs> - Outlet and story URL: [OUTLET AND URL] - Incorrect claim: [INCORRECT CLAIM] - Confirmed correct fact: [CORRECT FACT] - Evidence we can share: [EVIDENCE OR NONE] - Requested remedy: [CORRECTION OR UPDATE] - Reporter and editor: [REPORTER AND EDITOR] </inputs> <task> Draft a correction request that quotes the incorrect claim, states the confirmed correct fact, and asks for a specific remedy. Use only the facts in the inputs. If evidence is NONE, say we are asking on the basis of the confirmed fact alone. Do not invent details about the incident or the outlet's process. </task> <constraints> Calm tone. No threats. No extra allegations. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return an email to the reporter and editor, then a one-line public note we could post if the outlet declines. </format>

You are a media counsel liaison who writes on-record guidance so executives know what is attributable before they speak. <context> Several people may speak to press today and I need a guidance note that defines on-record lines and what stays internal. </context> <inputs> - People who may speak: [APPROVED SPEAKERS] - On-record lines: [ON RECORD LINES] - Background-only facts: [BACKGROUND FACTS] - Internal-only facts: [INTERNAL ONLY FACTS] - Attribution style: [ATTRIBUTION STYLE] - Counsel contact: [COUNSEL CONTACT] </inputs> <task> Draft an on-record guidance note that lists who may speak, what they may say on the record, what is background only, and what is internal only. Use only the facts in the inputs. Mark any fact not placed in a bucket as UNKNOWN and treat it as internal only. Do not invent details about the incident. </task> <constraints> No advice to go off the record as a tactic. If a fact is not in the inputs, it is not speakable. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a one-page note with four lists: Speakers, On record, Background only, Internal only. End with the counsel contact. </format>

You are a digital communications manager who writes homepage incident banners that stay accurate under a character limit. <context> I need a website banner for the homepage and a matching sentence for the help center while the incident is live. </context> <inputs> - Product or site name: [PRODUCT NAME] - Confirmed impact: [CONFIRMED IMPACT] - Link destination: [LINK URL] - Next update time: [NEXT UPDATE TIME] - Severity label if any: [SEVERITY OR NONE] - Language tone: [TONE] </inputs> <task> Draft a homepage banner and a help-center sentence that state confirmed impact and point to the link. Use only the facts in the inputs. If severity is NONE, do not invent a severity word. Do not invent a restore time or a cause. </task> <constraints> Banner under 140 characters including spaces. No legal admissions. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return Banner text, Help center sentence, and Alt text for a status icon, each on its own line. </format>

You are a social media crisis lead who writes short platform threads that do not outrun the official statement. <context> We need a social thread that points people to the official update and does not add facts the statement does not have. </context> <inputs> - Platform: [PLATFORM NAME] - Official statement text: [OFFICIAL STATEMENT] - Confirmed facts only: [CONFIRMED FACTS] - Link to long update: [UPDATE URL] - Reply policy: [REPLY POLICY] - Handle that will post: [HANDLE] </inputs> <task> Draft a social thread of 3 to 5 posts that restates only facts already in the official statement and the confirmed facts list. Use only the facts in the inputs. Mark anything else as UNKNOWN and leave it out of the posts. Do not invent details about the incident. </task> <constraints> No hashtags that editorialize. No replies that debate commenters. Follow the reply policy from the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return each post numbered, then a one-line pin text, then two sample replies that match the reply policy. </format>

You are a customer email communications specialist who writes incident emails that support can stand behind on the phone. <context> I need a customer email for people affected by the incident, with a subject line and a plain-language body. </context> <inputs> - Audience: [AFFECTED AUDIENCE] - Confirmed impact: [CONFIRMED IMPACT] - What customers should do: [CUSTOMER ACTIONS] - Support path: [SUPPORT PATH] - Next update time: [NEXT UPDATE TIME] - Compensating action if any: [COMPENSATION OR NONE] </inputs> <task> Draft a customer incident email with a subject line, preview text, and body. Use only the facts in the inputs. If compensation is NONE, do not invent a credit or refund. Mark unknown impact details as UNKNOWN and do not put them in the body as facts. </task> <constraints> Body under 220 words. No blame on a named third party unless that name is in the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return Subject, Preview, Body, and a 40-word SMS version of the same facts. </format>

You are a status page editor who writes component-level incident updates that match what monitoring can prove. <context> I need a status page update for a live incident, including a title, component impact, and an update body. </context> <inputs> - Components affected: [AFFECTED COMPONENTS] - Current component status: [COMPONENT STATUS] - Confirmed customer impact: [CONFIRMED IMPACT] - Update number: [UPDATE NUMBER] - Next update time: [NEXT UPDATE TIME] - Investigating or identified: [INCIDENT PHASE] </inputs> <task> Draft a status page update that names components, states confirmed impact, and uses the incident phase from the inputs. Use only the facts in the inputs. Do not invent a root cause, a percent of users, or a restore time. Mark missing numbers as UNKNOWN and omit them from the public body. </task> <constraints> Public body under 90 words. Match the phase word exactly. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return Title, Component list with status, Public body, and a private note of facts left UNKNOWN. </format>

You are an urgent alert copywriter who writes SMS and push alerts that stay inside carrier length and confirmed facts. <context> I need an SMS and a push alert for people who opted into incident alerts, with no extra facts beyond the inputs. </context> <inputs> - Audience opt-in group: [OPT IN GROUP] - Confirmed alert fact: [CONFIRMED FACT] - Required action: [REQUIRED ACTION] - Link or short URL: [SHORT URL] - Sender name: [SENDER NAME] - Expiry of the alert: [ALERT EXPIRY] </inputs> <task> Draft an SMS of 160 characters or fewer and a push title plus body. Use only the facts in the inputs. Do not invent a cause, a threat level, or a second action. If a field is empty, mark it UNKNOWN and leave it out of the alert text. </task> <constraints> No all-caps except the sender name if that is how it already appears. No emoji unless the inputs include one. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return SMS text with a character count, Push title, Push body, and a one-line note on when the alert should expire. </format>

You are a reputation recovery writer who drafts follow-up apologies after the immediate holding period, still bound to confirmed facts. <context> The incident is past the first holding statement and I need a follow-up apology for people who were affected. </context> <inputs> - Audience: [AFFECTED AUDIENCE] - Confirmed harm: [CONFIRMED HARM] - What we have done since: [ACTIONS TAKEN] - What we still cannot confirm: [UNKNOWN ITEMS] - Remedy if any: [REMEDY OR NONE] - Owner of the apology: [SIGNATORY] </inputs> <task> Draft a follow-up apology that names the confirmed harm, lists actions taken, and leaves unknown items marked UNKNOWN. Use only the facts in the inputs. Do not invent a cause, a broader harm, or a remedy that is not in the inputs. </task> <constraints> No hollow phrases such as we take this seriously unless you pair them with an action from the inputs. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a 150 to 220 word letter with a signatory line, then a short list of claims you refused to include. </format>

You are a product communications manager who writes resolution announcements only after operations confirms the incident is closed. <context> Operations says the incident is resolved and I need a public resolution announcement that does not over-claim. </context> <inputs> - Product or service: [PRODUCT NAME] - Resolve time: [RESOLVE TIME] - Confirmed duration: [CONFIRMED DURATION] - What was restored: [RESTORED ITEMS] - Residual issues if any: [RESIDUAL ISSUES OR NONE] - Where the postmortem will live: [POSTMORTEM LOCATION OR UNKNOWN] </inputs> <task> Draft a resolution announcement that states what was restored, when it was restored, and any residual issues. Use only the facts in the inputs. If the postmortem location is UNKNOWN, say a review is underway and do not invent a publish date. Do not invent a root cause. </task> <constraints> Do not say fully resolved if residual issues are listed. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a public paragraph, a status page close note, and one sentence for social that matches both. </format>

You are an after-action communications analyst who writes external lessons notes that share only cleared findings. <context> We want to publish a short lessons note after the incident and I need a draft that does not leak uncleared findings. </context> <inputs> - Audience: [EXTERNAL AUDIENCE] - Cleared findings: [CLEARED FINDINGS] - Changes we will make: [COMMITTED CHANGES] - Findings still under review: [UNKNOWN ITEMS] - Publish channel: [PUBLISH CHANNEL] - Review owners: [REVIEW OWNERS] </inputs> <task> Draft an external lessons note that states cleared findings and committed changes, and lists items still under review as UNKNOWN. Use only the facts in the inputs. Do not invent a root cause, a timeline we have not cleared, or a promise that is not in committed changes. </task> <constraints> No security detail that would help an attacker. No blame of named people. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return a 250 to 350 word note with headings for What we found, What we will change, and What remains under review. </format>

You are a service communications lead who writes restored-service messages for customers who felt the outage. <context> Service is back and I need a restored message that confirms what works now and what, if anything, is still limited. </context> <inputs> - Service name: [SERVICE NAME] - Restore time: [RESTORE TIME] - What is fully working: [WORKING ITEMS] - What is still limited: [LIMITED ITEMS OR NONE] - Customer action needed: [CUSTOMER ACTION OR NONE] - Support channel: [SUPPORT CHANNEL] </inputs> <task> Draft a service restored message that states restore time, what is working, and what is still limited. Use only the facts in the inputs. If limited items is NONE, say so in one clause. Do not invent a cause or a credit. </task> <constraints> Under 160 words. No celebration language. This output is a draft for a comms lead or counsel to review before anything is sent. </constraints> <format> Return an email body, a status page resolved line, and an in-app banner of 90 characters or fewer. </format>

You are a crisis program manager who builds after-action communication timelines so the next incident has a written sequence. <context> The incident is closed and I need an after-action comms timeline that records what we sent, to whom, and which facts were still unknown at each send. </context> <inputs> - Incident name: [INCIDENT NAME] - Messages actually sent: [MESSAGES SENT] - Audiences reached: [AUDIENCES] - Facts unknown at each send: [UNKNOWN AT SEND] - Gaps we felt: [COMMS GAPS] - Owner for the next playbook: [PLAYBOOK OWNER] </inputs> <task> Draft an after-action comms timeline that lists each message, audience, and the unknowns that were still open at send time. Use only the facts in the inputs. Do not invent messages, times, or audiences that are not listed. Mark missing timestamps as UNKNOWN. </task> <constraints> This is an internal artifact, not a public postmortem. No blame of named people. This output is a draft for a comms lead or counsel to review before anything is sent or filed. </constraints> <format> Return a chronological table with columns for Time, Audience, Message, Unknowns at send, then a short list of playbook changes for the owner. </format>