From Brand Facts to Caption Drafts — Brief, Voice, Then Line-Edits

How to use this collection

These three assets are plain-text prompts. Paste them into any ordinary AI chat that accepts text; no account linking, terminal, API key or installation is required, and none of them posts, schedules, sends or publishes anything. Nothing here was executed during this write-up. **Start with the goal before you write a single caption.** Paste <https://tokrepo.com/asset/6caf3434-c678-4ea5-83f7-a44d22002366> (Turn Goals and Brand Facts Into a Creative Brief) together with your customer goal, your approved brand information block, and any background notes. It returns a nine-section brief — Goal, Audience, Approved Facts, Tone and Voice Rules, Do-Not-Use, Message Pillars, Suggested Deliverables, Open Questions, Flags — where every pillar must trace to an approved fact and anything unsupplied lands in Open Questions. Answer those questions before moving on. **Prerequisite check:** if the goal is missing it stops and asks; if the brand block is missing it returns only Goal, Open Questions and Flags. Keep the flag list beside you for the rest of the pipeline, because the two later prompts inherit the same do-not-use words. **Then build reusable caption instructions.** Paste <https://tokrepo.com/asset/6b2a861b-503a-4c2c-b38e-c249b18973f9> (Brand-Voice Caption Planner) with the same brand brief. It produces five fragments (A–E), each with angle name, a paste-ready instruction block, required inputs, a length guide, a 5–10 item do-not-use list and one verification note, plus a banned-word rationale table and a pass/fail checklist. It flags contradictions in your voice note rather than guessing. **Then turn real notes into publishable drafts.** Paste <https://tokrepo.com/asset/3128c956-a9b4-41a8-8d39-6e7dec2bc503> (Reusable Draft-to-Final Writing Prompt) with five blocks — task, audience and relationship, tone, constraints, rough notes. It replies under Brief Check, Goal, Outline, Draft, Review, Variations, Next Step For Me, and its Review section names weak spots instead of praising. Vague notes come back marked [NEEDS DETAIL: ...]. **Practical progression:** brief (decide what is true) → fragments (decide how to say it) → draft (say it well) → you edit and send. An illustrative, fictional example: a small roaster with an approved brand block that allows the words "slow" and "small batch" and forbids "artisan" would first confirm its roast dates in the brief's Approved Facts, then generate a Fragment A for a first pour-over caption, then run the messy caption notes through the draft prompt — checking before publishing that no banned word survived. **Honest limitations:** all three are drafting aids, not guarantees. They can only see what you paste, so strip names, client figures and confidential material first. They do not verify claims, access live systems, or send anything; every fact, price and date must be checked against your own source. They deliberately avoid medical, legal and financial advice. Review is an instruction to the model, not an independent audit — the final editorial call stays with a human. Each source notes that it was source-reviewed but not runtime-tested.