Skip to content

docs: add weekly digest agent skill blog post#2673

Open
Lotte-Verheyden wants to merge 2 commits intomainfrom
blog-weekly-digest-skill
Open

docs: add weekly digest agent skill blog post#2673
Lotte-Verheyden wants to merge 2 commits intomainfrom
blog-weekly-digest-skill

Conversation

@Lotte-Verheyden
Copy link
Member

Adds the blog post "How We Built an Agent Skill to Synthesize 500+ Weekly Support Signals".

The post covers:

  • How we built an agent skill that synthesizes Plain tickets, GitHub issues, and Circleback meeting notes into a weekly digest
  • The iteration loop using Langfuse trace comments to improve the skill across 9 runs
  • Eval setup with LLM-as-judge evaluators (actionability, cites-sources, categories)
  • Lessons learned that generalize to other agents

TODOs before publishing

  • Add screenshot of example digest output (first section)
  • Add before/after skill output excerpts to the expandable sections (or confirm current excerpts are final)
  • Add the skill and scripts to the langfuse-examples repo and verify the link

@vercel
Copy link

vercel bot commented Mar 18, 2026

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
langfuse-docs Ready Ready Preview, Comment Mar 19, 2026 0:46am

Request Review

@github-actions
Copy link

github-actions bot commented Mar 18, 2026

📦 Next.js Bundle Analysis for langfuse-docs

This analysis was generated by the Next.js Bundle Analysis action. 🤖

New Page Added

The following page was added to the bundle from the code in this PR:

Page Size (compressed) First Load % of Budget (350 KB)
/blog/2026-03-13-weekly-digest-ai-skill 66.05 KB 477.38 KB 136.40%

@dosubot dosubot bot added the size:L This PR changes 100-499 lines, ignoring generated files. label Mar 19, 2026
import { FileTree } from "nextra/components";

<BlogHeader
title="How We Built an Agent to Synthesize our User Inputs"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would call it agent skill

Myb also: How We Built an Agent Skill to Synthesize what Langfuse Users want

authors={["lotteverheyden"]}
/>

At Langfuse, support tickets and GitHub issues go directly to the engineer who owns that area. Meeting transcriptions get shared around. People notice patterns and flag them in Slack. It works, but nobody has time to read across all of it every week. So we built an AI agent skill that does it for us: it reads through our support tickets, GitHub issues, and meeting transcriptions and synthesizes them into a weekly digest of what users are experiencing.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would maybe intro with: At Langfuse user inputs land on our desks through a variety of channels. Support tickets and Github issues go directly to engineers. Meeting notes from customer interactions get shared around. People notice ....

It works, but nobody has time to consolidate what's on our customers mind in one single source of truth every week. So we built an AI Agent ...


### Category names carry implicit assumptions

The third version introduced a "Quick Fixes" category, but AI doesn't know how quick a fix will be. We renamed it to reflect what we *do* know: whether the path forward is clear or needs discussion. Watch for labels that embed assumptions the model can't verify.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We could add a 'where we see it going section'

  • The skill auto assigning respective engineers
  • Incorporating more sources where langfuse is discussed
  • The skill based on documentation and codebase proposing approaches on how to deal with it,...

Copy link
Contributor

@annabellscha annabellscha left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added some comments! But i think its a nice illustrative use case!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

size:L This PR changes 100-499 lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants