Canned Response Writer

What it does

Looks at a set of recent chat transcripts or a list of frequently asked questions, groups them by intent, and writes a short, natural-sounding canned response for each that an agent can send with one click or light editing. Keeps each response generic enough to reuse but specific enough to actually answer the question, rather than vague filler.

Install

This skill isn't packaged in an installable registry. It's a plain SKILL.md file you can copy directly. Copy it from the SKILL.md tab and save it at the path for your assistant:

Use it in your AI assistant

Claude Code

.claude/skills/canned-response-writer/ (or ~/.claude/skills/canned-response-writer/ for all projects)

GitHub Copilot (VS Code)

.github/skills/canned-response-writer/

OpenAI Codex

.codex/skills/canned-response-writer/

Cursor

.cursor/skills/canned-response-writer/ (or ~/.cursor/skills/canned-response-writer/ globally)

Compatibility

Any WordPress site running a live chat plugin that supports saved or canned replies; also works from a plain list of FAQs with no chat plugin at all.

The complete SKILL.md for this skill:

---
name: canned-response-writer
description: Use when a user wants saved or canned chat replies drafted for common questions, or asks for "quick reply" templates or a "saved responses" list.
---

# Canned Response Writer

## When to use
Live chat agents keep retyping the same answers, and the user wants a set of ready-to-send canned responses for the questions that come up most.

## Procedure
1. Collect recent chat transcripts or an FAQ list and group questions into distinct intents (shipping time, refund policy, password reset, etc.).
2. For each intent, write one canned response in a natural, first-person support voice, no more than 3-4 sentences.
3. Include a placeholder in brackets, like [order number] or [customer name], wherever the reply needs personalizing before sending.
4. Order responses by how frequently the intent came up, most common first.
5. Flag any question that came up often but has no clear policy answer yet rather than inventing a policy.

## Guardrails
- Don't state a policy (refund window, shipping cost) that isn't confirmed in the source material, flag it as unknown instead.
- Keep tone consistent with the site's existing support responses, don't introduce a different voice.