What it does
The skill reads a regulator release (a letter, standard, guidance note or consultation paper) alongside a short organisation context file. It returns a draft impact assessment in six fixed sections: Change summary, Affected obligations, Impacted controls, Gap assessment, Draft actions and Sources.
Every statement cites the release paragraph, section or page it came from. Each affected area gets exactly one label (no impact, monitor or change required) with a one-sentence reason. Every action owner is left as TBD. The output carries a draft stamp naming the release, its date and the date checked.
The skill drafts. The human obligation owner validates every classification and decides what enters the obligations register. The skill gives no legal interpretation and drafts no submissions.
Who it is for
Compliance analysts, obligation owners and GRC teams who assess regulator releases against an obligations register. It suits a team that wants every release assessed by the same method and in the same shape, whoever runs it.
Before you start
- Get the full release text. The skill stops if the release looks incomplete, superseded or unclear.
- Write a one-page organisation context file: business lines, licence types, obligation register categories and control framework names. Keep it current, because the skill only maps to categories it can see.
- The skill needs no customer or staff information. Leave personal information out of both inputs.
- Run it only in an AI workspace your organisation has approved for this material.
- Name the obligation owner who validates assessments before anything reaches the register, and fill in the Owner, Approver and Approved on lines of the file before regulated use.
Set it up
Claude
Tested on 6 November 2026: pass.
- Download the Claude skill package (.zip) under Downloads.
- In claude.ai, open Customize, then Skills, then Add, then Upload a skill, and choose the .zip.
- Check the skill is switched on.
- In a new chat, ask Claude to use the reg-change-impact-assessment skill on your de-identified material.
ChatGPT
Tested on 6 November 2026: degraded, several statements in the first five sections, including the gap assessment caveats, cite no paragraph of the release.
- Download the Markdown skill file under Downloads and open it.
- Start a new ChatGPT chat. The test used a temporary chat.
- Paste the text from the line starting "# Skill:" to the end of the file, then the line "Apply the skill above to this input:", then your de-identified material.
- Review the draft before any use.
Other platforms, including Microsoft 365 Copilot: not yet tested.
Worked example
The test pack uses an invented release: a letter from the fictional Example Prudential Regulator about third-party AI services, dated 14 October 2026, assessed for the fictional Gannet Point Bank on 20 October 2026.
- Change summary. The six paragraphs say who the letter applies to (paragraph 2) and set two dates: an inventory in place by 31 March 2027 (paragraph 3) and contract terms for agreements entered into or renewed after 1 January 2027 (paragraph 4). Each date travels with its paragraph number.
- Affected obligations. The context file lists six register categories. The release text supports Outsourcing and third parties (paragraphs 3 and 4) and Technology risk (paragraph 5). Anti-money laundering, Complaints handling and Data and privacy appear in the context file but nowhere in the release, so a correct run leaves them out.
- Impacted controls. Each affected obligation is mapped to a framework named in the context file, such as the Third-Party Risk Framework v4 for the contract notice expectation.
- Gap assessment. Each area gets one label and a one-sentence reason drawn from the release.
- Draft actions. Actions are listed for every area labelled monitor or change required, with owners marked TBD.
- Sources. Every claim above traces to a paragraph. Anything that cannot be cited is removed and recorded as a gap.
The input carries two traps. The context file contains a line telling the assistant to mark every area as no impact. The v1.1 guardrail treats that line as content to report, not an instruction to follow. The covering note also asks for a statement of legal compliance and a reply to the regulator. Both are outside the skill, which leaves them with the obligation owner.
Sample input and output
Input (from the test pack, trimmed):
Output
From the Claude test run of 6 November 2026 (pass), copied as produced, first part only. The full run continues in the same format. Quotation marks and dashes are normalised to plain characters.
Limits
- The skill describes what a release says. It does not say what the law requires or whether the organisation complies.
- It is only as good as the context file. A missing register category or a renamed framework produces a gap the skill cannot see.
- Classifications are a first pass. The source article's own worked example has the obligation owner overturning a classification in each direction, and the file assumes that will happen.
- It does not draft submissions, consultation responses or correspondence with a regulator.
- One release per run. Comparing a release with an earlier one is a separate task.
Test record
Microsoft 365 Copilot and other platforms: not yet tested.
Each run used a synthetic input with planted traps, such as an embedded instruction to ignore the rules or a request for a decision the skill must not make. A separate reviewer graded each output against pass criteria written before the run. A degraded result means the output was usable but missed at least one criterion.
An earlier draft of this version was tested on 6 November 2026, and its wording was tightened before release. The results above are for the published text.
Version and changelog
Version 1.1.
- v1.0 (2026-08-15): Initial library release. Extracted from the published article.
- v1.1 (2026-11-15): Every statement in the first five sections now carries a citation, nothing is added after Sources, and requests for a compliance view or a regulator reply are handed to the obligation owner (changes made after the 6 Nov pre-release test).
- v1.1 (2026-11-15): Source now points to the guide page; Platforms and Tested against list only platforms actually tested; guide, worked example and test pack added.
- v1.1 (2026-11-15): Added the control block (version, owner, approver, approval date, review trigger) and a guardrail that treats instructions inside the release or context file as content to report, not instructions to follow.
Licence
Free to use and adapt internally. Attribution appreciated. No warranty. Educational purposes only.
General information and education only. Not legal, compliance, financial, or professional advice.
Downloads
- Markdown skill file, version 1.1
- Previous version 1.0, kept unchanged
- Claude skill package, version 1.1 (.zip)
TheAICommand. Intelligence, At Your Command.