Sephir recipe · meetings

Triage a long Slack thread before joining

Target query: summarize slack thread ai

The job

You join a channel after focused work and find a long Slack thread already in motion. Before you reply, you need the real state: what failed, what decisions were made, who owns each next step, and which blockers are still open. Scrolling message by message burns time and increases the chance you repeat solved points.

Why this is hard without Sephir

Slack threads mix decisions, questions, logs, and side chatter in one stream. Manual catch-up means scanning dozens of replies, then rewriting the timeline yourself. Copy-pasting into a separate chat window strips context and breaks flow, and you still have to verify who said what and when. The result is slow, error-prone, and easy to misread under deadline pressure.

How Sephir does it

  1. Open the exact Slack thread you need to triage in your current tab.
  2. Open Sephir in the sidepanel with Cmd+Shift+S.
  3. Ask for a thread triage with decisions, owners, blockers, and unresolved questions — or run /thread-tldr if you already saved it.
  4. Watch the audit timeline as Sephir runs extractPageText(tab) on the thread.
  5. Review the structured output in-panel and confirm missing or ambiguous items with a follow-up turn.
  6. Export the result as Markdown for your incident note, standup prep, or handoff doc.

The skill behind it

This recipe maps to a saved skill that turns long-thread catch-up into one command. It reads the thread text, then returns a clean operational summary you can act on immediately.

yaml
skill: /thread-tldrtools: extractPageTextintent: Read a long Slack thread and extract decisions, owners, blockers, and open questions.

What it costs

Sephir runs this on your own ChatGPT Plus (Codex OAuth) or API key. Typical usage is ~3,000–8,000 input + ~500–1,500 output tokens on Claude Opus 4, GPT-5, or Gemini 3 Pro. Single-turn runs fit Free; multi-turn follow-ups use Pro features. See .

Related

Try it free

Free tier covers BYOK + single-turn chat.