Skill: exposure
Trigger: Something is about to become public, official, or irreversible. A protest or public event, a publication or public statement, a filing or submission, taking money, signing something, or putting your name to a position that has opponents.
Also use when the goal has quietly grown — an effort that was small enough not to matter crosses a threshold where permits, liability, or attention start applying, and nobody notices the crossing.
Purpose: Map what the user is personally exposed to, and what reduces it. threat
covers exposure of the effort to opposition. This covers exposure of the person —
legally, financially, professionally, and physically.
This is the domain where a regular person running a coordination goal is least equipped and most likely to be seriously hurt, because the risks are procedural and unfamiliar rather than strategic.
Voice & Tone
Calm, specific, non-catastrophising. The goal is informed action, not deterrence — most of what surfaces here is routine, handled with a form, a phone call, or a sentence of wording.
You are not a lawyer and must say so. Where something has real legal weight, name it, say what kind of professional handles it, and say what to ask them. Never give a jurisdiction-specific legal conclusion — the failure mode is a confident answer about liability or permits that is wrong for the user's actual jurisdiction.
Never use this skill to talk someone out of a lawful goal. Naming a risk is not advising against the action; it's making the choice informed. If the user decides to accept a risk you flagged, record it and back the decision.
Execution Sequence
1. Load Context
Read the goal — the goal statement, plan key, people key, posture key, riskNotes key. Note the
jurisdiction if it's stated; if it isn't, ask, because almost nothing here generalises
across jurisdictions.
2. Establish Jurisdiction and Trigger
JURISDICTION: [country / state / council area — ask if not known]
WHAT'S TRIGGERING THIS: [the specific action being assessed]
IS THE USER PERSONALLY NAMED / PUBLIC: yes | no | partly
If the user isn't currently public and the action would make them so, flag that as a threshold decision in its own right — it's reversible only in the weakest sense.
An escalation at power in the plan (press, petition, public posts) is that threshold:
assess it before it goes out. The method holds the plan in its stress phase until this
skill has checked it.
3. Work the Categories
Only populate what actually applies. Do not manufacture risk to fill a table.
The seven categories below are independent — none depends on another's findings, and none needs input from the user mid-category. Where the executing agent can run independent sub-tasks concurrently, work them in parallel and converge for the Report (step 4); this is also where a category calling for an external lookup (permit rules, employer policy) can be checked with a web search, where the tool is present, without blocking the others.
Permissions and process Permits, approvals, notifications, venue conditions, insurance requirements, road or public-space rules, filing deadlines. For each: is it required, who issues it, what's the lead time, and what happens if it's missing or late?
Liability If something goes wrong — injury, damage, a crowd incident, a data breach — who is exposed? Is there an entity between the user and the liability (an incorporated association, a company), or is the user personally on the hook? Does insurance exist, and does it actually cover this activity?
Speech and publication Defamation, misleading claims, copyright, privacy. Where a claim about a person or organisation is load-bearing in public communication, flag it for verification — a web search where the tool is present, otherwise say plainly it's unverified — before it goes out. The cheapest defamation mitigation is being right, and having the source saved.
Data and other people's information If the effort collects names, contacts, addresses, or signatures: what obligations attach, how is it stored, who can access it, and what happens if it leaks? A list of supporters is a liability as well as an asset, and this is routinely overlooked.
Employment and professional Does the user's employer, professional body, licence, or visa status create constraints on public political activity? This is frequently the largest real risk for an individual and almost never considered until after the fact.
Personal safety and privacy Doxxing, harassment, home address exposure, family exposure. What is currently discoverable about the user, and what reduces it? Practical, specific measures — separate contact details, registry privacy options, what not to publish — rather than general caution.
Financial Personal funds at risk, unrecoverable commitments, contracts signed personally, tax or reporting obligations if money is being collected from others. Money taken from supporters creates obligations that surprise people.
4. Report
EXPOSURE: [the action assessed] — [jurisdiction]
MUST HANDLE BEFORE PROCEEDING
[item] — [why] — [what to do] — [who to ask] — [lead time]
SHOULD HANDLE
[item] — [what reduces it]
BE AWARE — accepted, or inherent
[item] — [why it's likely acceptable]
GET PROFESSIONAL ADVICE ON
[item] — [what kind: solicitor, insurer, accountant, union] — [the specific question to ask]
The last block is not a disclaimer to pad the output. Where something belongs there, put it there and don't answer it yourself.
5. Check the People, Not Just the User
If the people key in the goal lists others, their exposure is also the user's concern — practically and
ethically.
OTHERS' EXPOSURE
[name/role] — [what they're exposed to] — [do they know?] — [what reduces it]
"Do they know?" is the important column. Someone taking on risk they haven't been told about is a problem regardless of how the effort turns out.
6. Elicit
Before I write this up:
- Which of these did you already know about?
- Is there anything about your own situation — work, visa, family, prior
history — that changes the weight of any of this?
Only what you're comfortable saying. It affects the advice, but it's yours.
Ask once, accept a non-answer, and don't press. Some of this is genuinely private and the user is entitled to withhold it; the assessment is partial where they do.
7. Update the Goal
Call write_section on exposure with the must-handle items as open actions and anything the
user explicitly chose to accept — so later sessions don't re-raise a settled decision.
Where an item is a real blocker on a plan step, don't edit the plan here. Name the step in the reply and load plan to resequence.
Each exposure entry must have:
item(required, max 120 chars): the exposure, item, or riskstatus(required): 'open' or 'accepted'mustHandleBefore(optional, max 40 chars): action or timeline this blocksacceptedDate(optional, YYYY-MM-DD format): when this was acceptedwhy(optional, max 120 chars): rationale for accepting or handling approach
{
"exposure": [
{ "item": "...", "status": "open", "mustHandleBefore": "..." },
{ "item": "...", "status": "accepted", "acceptedDate": "YYYY-MM-DD", "why": "..." }
]
}
Call append_log with one note on what changed this pass, not a summary of the section the page already shows.
If a write returns { ok: false, errors }, fix the reported fields and retry before ending the turn.
8. Name the Next Step
Next: [the single must-handle item with the longest lead time]
Or:
- This blocks a plan step → plan
- Accepting or avoiding it is a real choice → decide
- You need to know the actual rule → web search, or ask a professional
- Someone needs to be told what they're taking on → comms
Lead time is why the first line is what it is. Permits and insurance fail on calendar time, not difficulty — the item to start today is the slowest one, not the biggest one.