Keep original meal descriptions and unknown portions
Provide fictional meal descriptions, dates, fields and food-label information. Draft a log and check duplicate or missing entries.
Execution environment and usage
AI / software
Text chat (ChatGPT, Claude or Gemini)
Where to operate
Optional text entry: start a new conversation at chatgpt.com; the supplied-text exercise can also be adapted to Claude or Gemini.
Getting started
- Start a new conversation at chatgpt.com and paste the original exercise prompt; do not use retired agent-mode onboarding.
- Provide fictional meal descriptions, dates, fields and food-label information.
- Review the requested result against its original inputs: Separate reported facts and food-label values from estimates. Do not perform actions excluded by the prompt.
Setup and plan requirements
Complete, authorized text/sample inputs specified in the prompt; no external account access required.
Prompt portability limits
Portable as a planning/drafting exercise; the historical product is provenance, not a necessary execution environment. This is an adapted exercise, not equivalent execution of the historical agent workflow.
Usage limitations
Current cross-surface Cowork, gradually unified into Claude; local and browser access still have prerequisites. Historical tool/source attribution is retained. Not hands-on tested; verify plan, region, workspace and site access separately.
Environment sources
適用情境 / When to use
Provide fictional meal descriptions, dates, fields and food-label information. Draft a log and check duplicate or missing entries. Do not invent numbers, set dietary targets or write directly to an external spreadsheet. Source record: Only the logging workflow is used, excluding the source’s nutrition targets, health advice and superiority claims.
步驟 / Steps
- Use fictional meal descriptions, field definitions and supplied label data.
- Preserve the original description and separate explicit portions from inferred ingredients.
- Check duplicates, dates and missing values, labeling estimate sources.
提示詞 / Prompt
From [fictional meal descriptions/date/field definitions/supplied food labels], draft a log. Preserve original wording and separate observations, label evidence and unknown portions. Do not treat missing logs as skipped meals or invent figures. Check data quality only; do not advise on weight loss, medication or nutrition targets.
留意事項 / Cautions
- Do not set calorie targets, diagnose or recommend supplements.
- Do not enter real medical or health records.
來源限制 / Evidence limits
Portions and ingredients may be estimates; an unlogged meal does not mean no food was eaten.
流程說明 / Workflow note
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
原始工具組合 / Original tool stack
Claude Cowork; Dispatch; spreadsheet
Source checked: 2026-10-20