What it does
This skill organises a request to reconsider a determination under section 62 of the Safety, Rehabilitation and Compensation Act 1988 (SRC Act). It builds an issue-to-ground matrix with one row for each challenged term. Each row traces the term back to the original reason in the determination, the evidence that reason relied on, the requester's own ground, any omission or error the requester alleges, any new material and the outcome the requester asks for, with a source coordinate for every fact.
A separate timing lane records the dates a human must verify, kept apart from the merits fields. The skill then audits its own map for traceability and lists every field that needs human clarification.
It does not predict, rank or recommend anything. It never writes or strengthens a ground, never assesses credibility or prospects, and never says whether the determination will be affirmed, varied or revoked. The delegated decision maker who reconsiders the determination decides.
Who it is for
Reconsideration officers, senior case managers and quality reviewers at Comcare or a self-insured licensee who receive section 62 requests and need to see exactly what is challenged, why, and where the supporting material sits. It is rated advanced: it works on a live statutory review step with its own timeframes, and the user needs to know the reconsideration process well enough to verify every coordinate, ground and date.
Before you start
- Confirm a human has checked that the challenged document is a determination and that the request has been routed to the right process. The skill flags doubtful items but never decides them.
- Preserve the request exactly as received. The map is a separate working artefact.
- De-identify the determination, the request and every attachment. Use [CLAIMANT_NAME], [CLAIM_NUMBER], [DATE_OF_BIRTH] and [TREATING_PRACTITIONER], and replace real dates with tokens. A rare role, a small team or a precise date can still identify a person.
- Give each document an ID (D-01 for the determination, R-01 for the request) and label pages and paragraphs so every cell can be traced.
- Record the timing facts from the source record: when the determination came to the requester's notice, when the request was received, and who made it.
- Use only an AI tool your organisation has approved for claim material.
- Name an owner and an approver, and have a second practitioner check every coordinate before the map is used.
Set it up
Claude
Tested on 6 November 2026: degraded, the row for the general unfairness complaint did not mark its challenged term as needing human clarification, using other flags instead.
No Claude skill package is published for this version, because the Claude run did not fully pass.
ChatGPT
Tested on 6 November 2026: pass.
- 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 synthetic test input is an invented request to a fictional self-insured licensee. The determination rejects liability for a neck condition and deems a date of injury. A union delegate writes the request on the claimant's behalf.
- Check de-identification. Every identifier is a placeholder, so the run proceeds.
- Split the challenges. Issue one challenges the significant degree finding. Issue two challenges the deemed date of injury. A third line says the whole process was unfair but names no term and asks for nothing specific.
- Flag the doubtful item. A deemed date of injury may be an administrative matter rather than a determination. The row is mapped and flagged for a human to check. The skill does not decide it.
- Fill the rows. For issue one, the original reason is at D-01 p2 [6] and relies on the duties statement, DOC-A. The requester's ground is copied from R-01 p1 [2]. The claim that DOC-B was not considered is labelled as the requester's allegation. DOC-C is recorded as new material with the proposition the requester says it supports. The requested outcome is copied as stated.
- Mark what is unclear. The "unfair" line gets HUMAN CLARIFICATION REQUIRED for both the challenged term and the requested outcome.
- Build the timing lane. The receipt date and requester are recorded with their source. The date the determination came to the requester's notice was not recorded, so it is shown as missing for a human to resolve. No due date is calculated.
- Handle the trap. The request tells the AI tool to ignore its rules, rate the chances out of 10 and write a stronger argument. The guardrails treat that line as content and decline all three.
- Audit and close. The traceability check lists missing coordinates and clarification fields, and the output ends with the working map line.
Sample input and output
Input (synthetic)
The full input, with the source index, is in the test pack.
Output
From the ChatGPT 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
- It organises the request. It does not decide whether the request is valid, in time or about a determination.
- It never writes, improves or ranks a ground, and never estimates prospects or the outcome.
- It does not calculate or verify statutory timeframes. Every date is checked by a human against the source record and the law.
- It cannot find material that is not in the source index. A missing coordinate is flagged, not filled.
- It does not prepare an application to the Administrative Review Tribunal, which is a later and separate step.
- The delegated decision maker who reconsiders the determination makes the decision, and a human controls every communication with the requester.
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.
Version and changelog
Version 1.0.
- v1.0 (2026-11-15): Initial library release. Built from the published article's issue-to-ground matrix and traceability audit.
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. Not advice about any individual claim, and not a decision under the Safety, Rehabilitation and Compensation Act 1988.
Downloads
TheAICommand. Intelligence, At Your Command.