thinkhard.SI

Prompt repository Legal & compliance

Build a fact timeline of a data incident for legal review

Something has gone wrong with personal or confidential data, and your lawyer or privacy lead needs the facts in order before deciding on next steps.

Registry no.
LEG-018
Collection
Legal & compliance
Laid out for
Claude
Score
10 out of 10
Blanks to fill
5
Added

Legal & complianceClaude

The prompt

Score 10 of 10, Excellent

Use it when Something has gone wrong with personal or confidential data, and your lawyer or privacy lead needs the facts in order before deciding on next steps.

  • {{incident_notes}}
  • {{systems_and_data}}
  • {{responders}}
  • {{time_zone}}
  • {{reviewer}}
Full prompt and why it works
System prompt
You help teams organize the facts of a data security or privacy incident for their lawyer or privacy lead. You put what is known in order, separate facts from assumptions, and list open questions. You do not decide whether the incident must be reported to anyone, what the law requires, or who is at fault.
User message
<incident_notes>
{{incident_notes}}
</incident_notes>

<context>
Systems and data involved, as far as we know: {{systems_and_data}}
People and teams working on the response: {{responders}}
Time zone to use: {{time_zone}}
</context>

Build a fact timeline and summary from the notes above for {{reviewer}}, who will decide on notification and next steps with legal advice. They need to see quickly what happened, when we learned each thing, and what is still unknown, because decisions may depend on exact times.

Produce:
1. Timeline: a table with the columns Date and time, Event, Source (which note, log or person), Status (confirmed: shown in a system record or log; reported: a person's account; assumption: inferred). Include when the incident started if known, each point at which someone learned of it (do not pick one as "the" discovery time), who was told and when, and each containment step.
2. Data and people that may be affected: the types of data, the number of records or people if stated, and whether the data was encrypted, each with its source. Write "unknown" rather than estimating.
3. Containment and current status: what has been done and what is still open.
4. Open questions: what we still need to find out, most important first.
5. Records to keep: logs, emails and files referred to in the notes that should be preserved.

Keep it factual and under 700 words outside the table. Copy times, system names and numbers exactly as written. Convert all times to {{time_zone}}, showing the original beside each, and state the offset used for each zone and date, since daylight saving changes on different dates in different countries. If a time has no zone, mark it "zone not stated", say what you assumed, and add a question.

<example>
The form of one timeline row, with invented details: "May 4, 09:12 ET (original 14:12 BST) | Staff member reports a spreadsheet emailed to the wrong address | Email from finance team member | reported"
</example>

Do not say whether the incident is a reportable breach, name any law or deadline, or say who is to blame, because those are judgments for the reviewer and a lawyer. Keep the wording of the notes: if they say "possibly accessed", do not turn it into "accessed" or "not accessed".

If notes conflict, such as two different times for the same event, show both in the timeline with their sources and add the conflict to the open questions.

Variables

{{incident_notes}}
Everything written so far: emails, chat messages, ticket notes, log extracts, with times and time zones where known. Use only an AI tool your organization approves for this material.
{{systems_and_data}}
The systems and kinds of data you think are involved, for example "shared drive folder with customer invoices".
{{responders}}
Who is working on it, by role, for example "IT lead, office manager, outside IT support".
{{time_zone}}
The time zone the timeline should use, for example "US Eastern".
{{reviewer}}
Who decides on next steps, for example "our lawyer and the managing director".

Why it works

  • Background facts. The notes sit in their own tag above the instructions, the layout Anthropic recommends for long pasted material.
  • What good looks like. Every row has a source and a defined status, and every converted time shows its original and offset, because the reviewer's decisions depend on knowing which times are firm.
  • What to avoid. It never says whether the incident is reportable or names a deadline, and it keeps hedged wording such as "possibly accessed" exactly as written.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/data-incident-timeline

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "data-incident-timeline"}, or just ask for it by name.