The poster wants a persistent Claude agent that watches Slack and call transcripts, spots tasks, and drafts the Claude Code prompts, removing them from the loop.
The poster wants a persistent Claude agent that watches Slack and call transcripts, spots tasks, and drafts the Claude Code prompts, removing them from the loop.
The post itself is a request, not a finished build: the poster says their company’s Claude enterprise setup already connects to Slack, call recordings/transcripts, a data warehouse, a knowledge base, Google Workspace, and internal tools, and they want a permanent agent to watch Slack and transcripts, identify tasks, draft Claude Code prompts, and let them approve or tweak before running. In comments, the most concrete similar setup described is a small always-on Slack app on a Mac that listens for mentions and alerts, starts Claude Code in headless mode per Slack thread, keeps phase one read-only, and gates write actions, sends, and higher-risk changes behind approval buttons.
No outcome is reported by the original poster because they are asking for advice. A commenter who built a similar system says it was a "total game changer," but gives no hard metrics; others warn quality can drop without strong approval, deduping, and read-only staging.
Most actionable catch: start read-only, separate task extraction from execution, and dedupe requests because the same ask may appear in both Slack and call transcripts.
Read the [Slack export or channel folder] and the call transcripts in [folder path], and find action items that could be handed to a coding agent for [project name]. For each one, draft a clear, ready-to-run Claude Code prompt and note which message or transcript it came from. Save them as a single list in [file path], and don't run any of the tasks or send any messages yourself.
Score every demo request against your ICP, update HubSpot and ping the right rep in Slack within a minute of the form submit.
Every new Zendesk ticket gets a category, priority and sentiment tag plus a draft reply in an internal note before an agent opens it.
Plan, execute and review a cross-cutting refactor in small batches, with Claude Code making the edits and GitHub Copilot reviewing each PR.