INPUT 01
Input 01
An approved brief and business goal
Referrals
A source-backed referrals workflow for agencies. Review its inputs, setup and geographic limits before your first draft.
Work typeMarketing
SetupInstruction-led
ContextWorldwide
01 / What you bring
REQUIRED INPUTSGather the material before running the prompt. Follow the publisher’s full checklist for the specific task.
INPUT 02
Input 02
Audience, brand examples, budget or source metrics
INPUT 03
Input 03
The complete skill folder and any required supporting files
Publisher excerpt: “When the user wants to create, optimize, or analyze a referral program, affiliate program, or word-of-mouth strategy. Also…” Source ↗
02 / What you get
EXPECTED OUTPUTA reviewable marketing plan, draft, or analysis
DRAFT HANDOVERFOR HUMAN REVIEW
A reviewable marketing plan, draft, or analysis
Evidence and supplied inputs
Assumptions and missing information
Draft output and review questions
This describes the intended handover. It is not a measured or executed result.
03 / Set it up
Claude: follow the publisher’s Claude or Claude Code installation path. Preserve the complete skill folder and supporting references; configure required tools before trying the prompt.
Codex: use the publisher’s Codex or Agent Skills setup path. Keep supporting scripts and references, inspect required permissions and validate a permitted sample.
ChatGPT: manual instruction adaptation. Read the source, supply the required context and files, and configure available tools yourself. This is not a native ChatGPT plugin installation.
STEP 01
Review the complete source
Review the linked source and the publisher’s current README before installing.
STEP 02
Keep the full workflow
Download or install the complete plugin/skill folder, preserving references and scripts. Do not paste a terminal command intended for another app.
STEP 03
Configure your workspace
Use the setup path for the platform you actually have; configure only the required apps and permissions.
STEP 04
Try a reviewable first draft
Start with a permitted sample and the starter prompt. Check the first result against your own normal review process.
Compatibility, tools & dependencies
Publisher documents Claude Code and Codex/Agent Skills setup. Product-specific tools and supporting files still need configuration. ChatGPT is a manual instruction/prompt adaptation only; no native plugin installation or publisher-tested ChatGPT execution is claimed.
Geography-neutral workflow. Set Australian spelling, AUD, dates and business context explicitly; local compliance is not supplied by this label.
Named tools: Stripe. Actual access depends on your setup.
- Instruction-led workflow. Preserve any references bundled beside SKILL.md.
Pinned source snapshot; review publisher release and installation notes.
Public instructions do not imply free access to AI products, apps or APIs.
04 / Your starter prompt
REPLACE THE BRACKETSUse referrals for [business task]. Our business is in [country and state/territory], our currency is [currency], and the intended reader is [audience]. Use only the attached authorised inputs and approved policies. First identify missing information and unavailable tools. Prepare a draft for human review; cite the supplied evidence, separate assumptions, and do not send, publish, lodge, purchase or change records.
Fieldwork’s starter prompt. Keep the complete publisher workflow and replace the brackets before use.
05 / Before you rely on it
Review the draft before using it
Use supportable claims and the business’s approved voice. Verify attribution and results, and get approval before publication or campaign changes.
This is a source-inspected starting point. Fieldwork has not run-tested this skill or measured time savings.