Select ChatGPT on the task page. Choose the working mode, audience and requested output, then inspect and copy the task starter.
Paste it into a chat. Supply only material you are authorized to share, following your organization's rules and your ChatGPT account controls. The starter asks for missing inputs before it drafts.
When you reach a stage, copy that stage's instructions. They include the actual source methods and named artifact requirements. Check the result before proceeding.
Keep each artifact in your own files. For revision, bring the prior version, issue list and accept/reject/amend decisions. To resume in a new chat, copy the resume instructions and bring the listed artifacts and methods again.
Use ChatGPT's actual upload and file controls if your account supports them. Ask it to confirm the filename, relevant pages or sheets and readable content before proceeding. A failed upload, unreadable image or missing sheet requires a new readable extract or complete text/tables. Do not treat unseen material as evidence.
Text, Markdown and tables are the default. Request DOCX or XLSX only if the current assistant can actually create an attachment and read it back. Check headings, tables, row counts and any formulas and displayed results. If unavailable, retain the same substantive text or tables with save steps. A filename in a message does not prove a file exists.
Account controls and evidence
This website does not run ChatGPT or receive your documents. Anything you provide in ChatGPT is processed under your own account settings, policies, workspace controls and limits. A copied method does not establish model accuracy, file support or a successful task run.
Earlier private review-adapter checks cover bounded English cases. They do not establish every production task, model, native output or current portable task route. Inspect the methods and verify outputs against your sources.
Optional skills and plugins
Portable copy instructions do not need an installed skill. OpenAI documents stand-alone skills on desktop and Codex; plugin-bundled skills have their own supported surfaces. Availability and organization controls can vary. This library does not provide a public installation link for the private review adapter.
