What it does
This skill turns a settled, de-identified safety learning into a short draft that workers will read: a safety alert, a toolbox talk introduction or a bulletin. Every draft has the same four parts: what happened, what it means on this site, what to do now, and who to contact. A toolbox intro adds two open discussion questions. Each channel has a word cap: 80 words for an alert, 120 for a toolbox intro and 200 for a bulletin.
It describes what happened, never who failed. Every draft ends with a release line, because nothing goes out until the health and safety representative (HSR), or the affected workers where there is no HSR, has been consulted and a competent person has approved it. Drafting is never incident notification and never delays a call to the regulator.
Who it is for
WHS advisers, site supervisors, team leaders and HSRs who write safety alerts and toolbox talks in Australian workplaces. It is rated foundation: the inputs are five plain fields and the output is short, and the main discipline it asks of the user is de-identification and settled facts.
Before you start
- Use a learning whose facts are settled. Unsettled facts belong in an investigation, not an alert.
- Describe people by role, never by name, and leave out any detail that could identify someone.
- Decide the one action workers need to take. If there are two, decide which leads before you start.
- Know who will answer questions, usually a supervisor or the HSR, and who approves release.
- Use only an AI tool your organisation has approved for safety information.
- Name an owner and an approver for the file inside your team.
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 safety-comms 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 fall from a step ladder in a loading dock at a fictional freight business, for a toolbox intro at two depots.
- Check de-identification. The storeperson and the team leader appear by role only, so the run proceeds.
- Confirm the one action. One action is supplied: check the ladder's feet and the floor under it before every climb. Had two been supplied, the skill would ask which leads.
- Draft from supplied facts. What happened draws on the review's two findings: worn ladder feet and a wet floor from rain under the roller door. Nothing is added about dates, other injuries or causes not in the input.
- Describe, do not blame. The team leader asks for a line saying the storeperson skipped the pre-use check. That would name a failure by a person, which the method and guardrails forbid, so it is left out.
- Keep to the cap. A toolbox intro has 120 words. The draft keeps to four short parts plus two discussion questions.
- Close. The draft ends with the release line, and goes to the HSR and a competent person before any worker sees it.
Sample input and output
Input (synthetic)
Output
From the Claude test run of 6 November 2026 (pass), copied as produced. Quotation marks and dashes are normalised to plain characters.
Limits
- It drafts from settled, de-identified facts only. It is not for live incidents or investigations.
- It is not incident notification and does not replace a call to the regulator.
- It cannot know site history the inputs leave out. The HSR often adds context that changes the draft.
- It does not decide what action workers take. The action comes from the people accountable for the control.
- The word caps keep drafts readable, but a short draft is not automatically a clear one. A person still checks it reads well on the site.
- Every output is a draft until HSR consultation and competent person approval.
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. Release guardrail now covers workplaces without an HSR, where the affected workers are consulted instead.
- 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.
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.