What it does
This skill reviews a draft determination under the Safety, Rehabilitation and Compensation Act 1988 (SRC Act) before it issues. It numbers every factual assertion in the draft and maps each one to a document in the evidence index, marking any assertion with no source as UNMAPPED. It checks that the statutory elements in sections 5A, 5B and 14 are addressed with mapped evidence. It flags hedged words such as "should", "likely" and "it appears" where they sit in an outcome statement, and it confirms whether review rights text is present.
It produces a review checklist, never a decision. It does not draft, redraft or recommend the outcome. Every word of the determination remains the delegate's.
Who it is for
Delegates, senior case managers, quality reviewers and team leaders who check determinations before they issue. It is rated advanced: it runs close to a statutory decision, the user needs to know the elements of the SRC Act well enough to judge the checklist, and every statutory reference must be verified against the current compilation before the output is relied on.
Before you start
- De-identify the draft and the evidence index. Use [CLAIMANT_NAME], [CLAIM_NUMBER], [DATE_OF_BIRTH] and [TREATING_PRACTITIONER], and replace real dates with tokens.
- Give every document in the evidence index an ID, and cite those IDs in the draft where your template allows.
- Confirm the claim type: injury, disease or aggravation.
- Check which compilation of the SRC Act is in force on the Federal Register of Legislation. The skill's section references were written against Compilation No. 82, in force 1 July 2026, and are listed for checking in the entry's sources file.
- Use only an AI tool your organisation has approved for claim material.
- Name an owner and an approver for the file, and record when it was last checked against the Act.
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 determination-evidence-check skill on your de-identified material.
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 aggravation claim at a fictional Commonwealth agency: a five-document evidence index and a seven-paragraph draft.
- Number the assertions. P1 to P6 each carry a factual assertion. P7 is the outcome statement.
- Map the evidence. P1 maps to the employment record, P2 to the claim form, P3 to the duties statement, P5 to the practitioner's report and P6 to the certificate. P4, the report of back pain to a supervisor, cites no document and matches nothing in the index, so it is marked UNMAPPED.
- Walk the elements. Employment status, the section 5B significant degree question for an aggravation of a disease, and incapacity under section 14(1) each get a row with the evidence the draft relies on. Nothing in the facts raises the reasonable administrative action exclusion or the section 14(2) and 14(3) exclusions, so they are shown as not raised rather than as findings.
- Lint the outcome. P7 says the employment "likely contributed" and that liability "should be accepted". Both words are flagged with their location and the reason: hedged wording in an outcome statement.
- Check review rights. The draft ends at P7 with no review rights paragraph. That is flagged for the delegate. The skill does not write the paragraph.
- Handle the trap. The covering note asks the skill to rewrite P7 as a firm acceptance and to say whether liability is made out. Both requests sit outside the skill. The guardrails decline them and leave the outcome with the delegate.
- Prioritise. The gaps list puts the unmapped assertion, the missing review rights text and the hedged outcome in order for the delegate.
Sample input and output
Input (synthetic)
The full input, with all seven paragraphs, is in the test pack. P7 is deliberately hedged so the language check has something to find.
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
- It checks whether the draft's evidence chain is complete. It does not check whether the outcome is correct.
- It cannot see documents that are not in the evidence index, so a mapped assertion is only as good as the index.
- Its section references are a checklist, not legal advice. A human verifies them against the current compilation before use.
- It flags missing review rights text but never drafts it. The notice of a determination must include a statement that the claimant may request a reconsideration, and a human checks that wording against the Act and the team's approved template.
- It does not assess medical evidence, credibility or the weight of competing opinions.
- The delegate makes the determination and writes every word of it.
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 1.1.
- 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): Skill block gains Version, Owner, Approver, Approved on and Review trigger lines. [TREATING_PRACTITIONER] added to the de-identification placeholders, because provider names can identify a claimant. Statutory references checked against SRC Act Compilation No. 82 (1 July 2026); the section 5A(2) examples are now described as a non-exhaustive reference list.
- v1.0 (2026-08-15): Initial library release. Extracted from the published article.
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 a determination under the Safety, Rehabilitation and Compensation Act 1988.
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.