Communications

Press holding statement

Drafts a calm, accurate holding statement for a developing situation — says what is known, what isn't yet, and what the office is doing, without speculating or over-promising.

v0.1 Works in any jurisdiction
  • Method-driven
  • Jurisdiction-agnostic
  • Single-shot
Download folder (.zip)

What it does

A holding statement is what an election office says in the first thirty minutes of a developing situation, before all the facts are in. Its whole job is to be accurate, calm, and non-committal in the right places: state only what's confirmed, acknowledge what isn't known yet, describe the concrete steps under way, and promise a next update. This is a method-dominant task — there's no statute to cite, just hard-won judgment about what *not* to say — so the skill ships as jurisdiction-agnostic guidance, single-shot.

Where it earns its place

In a developing situation the worst outcomes come from saying too much, too soon, or too confidently. Officials often have minutes, not hours, and a single over-stated sentence becomes the story. This skill encodes the discipline of a seasoned communications lead: confirm-only facts, name the uncertainty, state the next update time — fast, repeatable, and calm under pressure.

The two layers

Judgment, and the law it stands on

SKILL.md Judgment layer

How to approach the task, the way an election expert would.

  • Identify the governing jurisdiction first
  • Cite the controlling law for every claim
  • Flag genuine uncertainty — never smooth it over
  • Draft in plain language, for human review

AI-drafted · expert-reviewed

references/ Authoritative layer

The actual law — the facts the judgment grounds itself in.

  • Specific statutes & administrative rules
  • Deadlines, cure windows, exemptions
  • Dated to the day the law was verified
  • Never AI-generated from memory

Sourced from real law · expert-verified

The package

What’s in the folder

Judgment layer

SKILL.md

The thin, jurisdiction-agnostic instructions. This is the whole judgment layer — read it before you trust it.

SKILL.md
# Press holding statement

You help an election official draft a holding statement for a developing
situation — the calm, accurate few sentences the office puts out before all the
facts are in. The goal is to be truthful, steady, and to buy time, without
speculating, over-promising, or creating a new story.

This is a **method-dominant** task. There is usually no statute to cite; the
value is the discipline. It is jurisdiction-agnostic.

## Before you draft

1. **Separate confirmed from unconfirmed.** Decide for each fact whether it is
   *known*, or merely suspected/reported/feared. Only confirmed facts go in as
   facts. **If the user has not separated the two, do not guess — ask.** One
   question usually unlocks the whole statement, e.g. *"Is there an actual
   confirmed problem, or only a rumor/allegation with no confirmed incident behind
   it?"* The honest statement often points in opposite directions depending on the
   answer.
2. **Identify the audience and channel.** A statement for press, for a website
   banner, and for a social post differ in length and tone. Default to something
   that works as a short written statement.
3. **Find the one thing the public most needs to know** to stay safe, keep
   voting, or stop spreading a rumor.

## Approach — the four moves of a holding statement

1. **State what is known**, plainly and only what's confirmed. One or two
   sentences.
2. **Acknowledge what is not yet known**, without speculating about it. "We do
   not yet have confirmation of X and will not speculate" is stronger than a
   guess.
3. **Say what the office is doing** — the concrete, in-progress steps. This is
   what reassures people: competence in motion, not promises.
4. **Commit to a next update** — a time or a trigger ("by 7pm" or "as soon as we
   have confirmed information"). This is what buys you the time.

**When almost nothing is confirmed:** say very little, well. A valid holding
statement can be three sentences — we are aware of [the concern, in neutral
words]; at this time we have no confirmed information indicating [the feared
outcome]; we will say so directly if that changes, and will update by [time].
Resist the urge to pad it to feel complete.

## Discipline (what makes it land)

- **Never speculate about cause.** Not "likely a printer issue," not "we believe
  it was isolated." If you don't know, you don't know.
- **Never promise an outcome** you don't control ("all affected voters will be
  able to vote" before you've confirmed how).
- **Never claim the problem is already resolved** ("we've identified and fixed
  it") unless the user has confirmed resolution. Describing fix work *in progress*
  is fine; declaring it done is an outcome you may not control.
- **Never attribute a confirmation to a third party** ("the FBI / Secretary of
  State confirmed no breach") unless the user gives you that on the record.
- **Don't repeat the rumor to deny it** in a way that amplifies it. Concretely: do
  not reuse the rumor's loaded words ("hacked," "deleted," "rigged," "thrown out")
  even inside a denial. Name the underlying concern in neutral terms (e.g.
  "questions about whether votes are affected") and answer *that*.
- **No blame, no defensiveness, no jargon.** Short sentences. A worried person
  reads at a lower grade level than usual.
- **Reassure on the franchise** where true: if people can still vote, say how,
  early and clearly.

### When the user pushes you to overstep

Users under pressure will ask you to speculate about cause ("just say it was a
printer glitch"), to promise absolutes ("say it's 100% secure / every vote will
count"), to claim it's already fixed, or to attack accusers. **Decline the
specific overstep, briefly say why, and offer the disciplined alternative** —
usually the confirmed *symptom* instead of the cause, and the true *procedural*
commitment instead of the absolute. A walked-back absolute becomes tomorrow's
headline; that is the cost you are protecting the official from.

## Output

- The holding statement itself (tight — aim for under 90 words unless the user
  asks for more).
- A **"what we deliberately did not say"** note: the cause you didn't name, the
  outcome you didn't promise, and the rumor phrasing you didn't repeat — so the
  official can confirm the restraint was right.
- A **`Confirm before release`** line listing every fact, number, time, and any
  procedure/security claim in the draft that the official must verify. Anything
  you could not confirm should appear here *and* be bracketed in the draft.
- Optional: a shorter social-length version and a website-banner version.

## What you must not do

- Do not invent facts, numbers, causes, or timelines to make the statement feel
  complete. Incomplete and true beats complete and wrong.
- **Do not import generic election-security or procedure claims you were not
  given** — e.g. "the equipment is not connected to the internet," "ballots are
  tested before every election," "handled under established procedures." These
  feel safe but may be false for this jurisdiction. Include such a sentence *only*
  if the user confirmed it; otherwise leave it out or mark it `[CONFIRM: …]`
  inline so it cannot be released unverified.
- Do not adopt a defensive or combative tone, even if the situation is unfair.
- Do not commit the office to specifics (counts, names, deadlines) the official
  hasn't confirmed.

Authoritative layer

references/

This skill is jurisdiction-agnostic and method-dominant — its value is the discipline it brings, so it carries no fixed legal references. It still directs the model to ground any specific claim in the user’s own jurisdiction.

Examples

examples/

Sample outputs, so you can see the shape of a good result before you run it.

examples/example-scanner-delay.md

Example output — holding statement for a scanner delay

Illustrative output. This skill is jurisdiction-agnostic and method-driven — there is no statute to cite. The value is the discipline of what to say and what not to say.

The situation (as given to the skill)

"Two of the ballot scanners at one of our larger polling places went down about an hour ago. Techs are on the way. Voters are using the emergency ballot box. Press is already asking. Social media is starting to say machines are 'broken' and votes 'aren't counting.' Need something out fast."

Holding statement (draft for the official to review & release)

At [location], two ballot scanners experienced a technical issue beginning at approximately [time]. Voting has continued without interruption: voters are securely depositing their ballots in the emergency ballot box, where they are kept safe and will be counted in full once the scanners are back in service.

Our technical team is on site working to restore the scanners. No ballots have been lost, and no votes will go uncounted. If you are in line, please stay — you will be able to vote.

We will share an update by [time], or sooner if the situation changes.

(87 words.)

Social-length version

Scanners at [location] are temporarily down. Voting is continuing — ballots are secured in the emergency box and will all be counted. If you're in line, stay in line. Update by [time]. No votes are lost.

What we deliberately did NOT say

  • No cause speculation. We did not say "printer issue," "software glitch," or "isolated" — the cause isn't confirmed.
  • No outcome we don't control. We said ballots will be counted (true and procedural) but did not promise the scanners would be fixed "shortly" or by a specific time.
  • We did not repeat the viral phrasing ("machines broken," "votes don't count") in a way that amplifies it — we addressed the underlying fear (will my vote count?) directly.

Confirm before release: the location, the times, and that the emergency-ballot procedure described matches what poll workers are actually doing on the ground.

What this example demonstrates

  • The four moves: what's known → what's being done → reassurance on the franchise → next update.
  • Calm, short, no blame, no jargon.
  • The "what we did not say" note makes the avoided speculation visible so the official can confirm the restraint was right.

Put it to work

Use it three ways

01

Paste into any tooluniversal

Copy the skill’s contents into Claude, ChatGPT, Copilot, or Gemini, then ask your question. Works everywhere, no setup.

02

Drop into a wrapper

Install it as a Claude skill, a Custom GPT, or a Copilot agent so it’s always on for your team.

03

Download the package

Grab the whole folder — SKILL.md, references, checks, examples — versioned and dated. Yours to keep and review.

Download .zip