← All use cases

中文

Automation & agents · Claude Code + Claude

Turn Slack requests into drafted tasks for a coding agent, with your approval

Idea + starter prompt, not run by us yet

For: Technical leads who have become the human go-between for requests scattered across chat and calls

This started as a Reddit question, not a finished build. The poster wants an always-on agent that keeps an eye on Slack and call transcripts, picks out tasks and writes draft prompts for Claude Code, which they approve or edit. A commenter described a similar setup: a small Slack app that starts Claude Code per thread, read-only at first, with approval buttons before any action.

Steps

  1. Start with read-only. The agent only reads Slack and transcripts and writes down proposed tasks.
  2. Separate finding tasks from doing them. Each proposed task is a draft prompt waiting for your approval.
  3. Remove duplicates, since the same request can show up in Slack and in a call transcript.
  4. Put sends, writes and risky changes behind an approval step, such as a button.
  5. Review the first week of drafts by hand before you let anything run.
  6. Without strong approval, dedupe and read-only staging, commenters warn that quality drops.

Starter prompt

Starter prompt. Change the [brackets] first.

I want to turn requests from Slack messages and call transcripts into task drafts for Claude Code. This stays read-only. It writes drafts in this chat and changes nothing anywhere else.

Trial first. Below are 2 made-up Slack messages and a short made-up call snippet that I will paste. Run the whole job on those only, so I can check the format. Do not read my real Slack or real transcripts until I say so.

For the real run, I will paste [Slack messages or transcript text, with names removed]. Find every request that needs engineering work. For each one, give me: a one-line title, who asked and where, what they want, a draft prompt for Claude Code that I could approve, and how risky it is.

Merge requests that look like duplicates and tell me which ones you merged. Do not run any task, post to Slack, change any file or send any message. Stop and wait for my yes before anything beyond drafting. Showing me a draft is not permission to run it. After I review the drafts, suggest how to move to a version where each task needs my click before it runs.

Original post: Reddit · found via agentusecases.ai. Written up in our own words. We haven't run it, so test it on made-up data first.