Customer support reply drafter grounded in your policy text
Drafts a support reply from a ticket, account context and the policy excerpts you paste in. It answers only what your policies allow, flags anything it cannot promise, and adds internal notes for the agent.
Prompt in Customer support by AISkillHub Library
What it does
Support replies go wrong in two ways: they promise something policy does not allow, or they are technically correct but cold. This prompt drafts a reply that is grounded only in the policy text you paste in, matches your tone, and separates what the customer sees from what the agent must do before sending.
**Use it when** you handle tickets by email or chat and want a solid first draft you can check and send, especially for billing, account, shipping, returns and "how do I" questions.
**What you get** - A one-line read of the ticket: category, what the customer wants, emotional temperature. - The reply itself: acknowledges the specific issue, answers or resolves it, states what happens next and when, and closes with one clear path if it does not work. - Internal notes: the actions the agent must take before sending (for example, issuing the refund the reply says has been issued), the policy lines relied on, and anything that needs a supervisor.
**What it will not do** - It will not invent policy, timelines, discounts, credits or compensation. If the policy excerpts do not cover the request, the reply says you are checking and the internal notes say what to escalate. - It will not admit legal liability or speculate about the cause of an incident beyond what the context states. - It will not ask the customer for passwords or full card numbers.
**Example** A frustrated customer reports a duplicate annual charge. With a policy excerpt allowing agent-issued refunds, the draft confirms the duplicate, says the refund has been issued, gives the realistic bank timeline from the policy, and tells the agent to issue the refund before sending. See the example file for the full run.
How to use it
## How to use
1. Open `prompts/support-reply-drafter.md` and copy everything below the title line. 2. Replace each `{{variable}}` with your own content. Paste long material (notes, transcripts, tables) directly in place of the variable; keep the surrounding tags so the model can tell inputs apart. 3. Start a new chat in ChatGPT, Claude or Gemini and paste the filled prompt as your first message. 4. Read the output against the quality checks at the end of the prompt before you use or send it.
## Variables
- `{{ticket}}`: the customer's message or the full thread, newest message last. - `{{customer_context}}`: plan, account age, order details, relevant logs or history. Remove data you would not want in a chat tool. - `{{policy_excerpts}}`: the exact policy, help-center or macro text that applies. The reply will not go beyond this. - `{{product_name}}`: your product or company name as customers know it. - `{{agent_name}}`: the name to sign with. - `{{tone}}`: e.g. `warm, direct`, `formal`, `friendly and brief`.
If you do not have a value for an optional variable, write `none` rather than deleting it, so the prompt still reads correctly.
## Example
`prompts/support-reply-drafter.example.md` shows the prompt with every variable filled and a trimmed sample of the output, so you can see the expected shape before you run it on real material.
## Reusing it
To reuse it often, save the prompt (with the variables left in) as the instructions of a Claude Project, a custom GPT or a Gemini Gem, then paste only the inputs in each new chat.
## Tips
- Keep a text file of your most-used policy excerpts so you can paste the relevant one quickly.
Files
Save each file at the path shown, relative to your project root.