blog-discourse

Blog Discourse is an API-free research skill that captures what practitioners and customers are actually saying about a topic on the public web over the last 30 (or 90) days. It solves the recency-and-engagement gap left by authority-first research: instead of citing established sources, it surveys live discussion across Reddit, Hacker News, X/Twitter, YouTube, dev.to, Medium, GitHub, Stack Overflow, and Substack. It produces DISCOURSE.md, a structured brief, plus JSON output a downstream writer can consume. The methodology is adapted from the last30days-skill by Matt Van Horn, but replaces platform APIs with WebSearch using platform-targeted site operators, so no API keys are required.

The workflow is staged. A mandatory topic pre-flight runs four keyword-trap checks and either reframes or asks a clarifying question before spending searches. Named-entity topics are decomposed into discrete queries covering primary entity, counter-perspective, practitioner discourse, tangential entities, and a time anchor. Phase 2 runs four to eight platform-targeted WebSearches with recency filters, and Phase 3 collects each result into a JSON record with platform, URL, title, snippet, date, and an engagement proxy. Notably, it writes results to a securely created temp file rather than a predictable path, and enforces an untrusted-data contract that scans snippets for prompt-injection patterns before they enter the pipeline.

It is aimed at content writers, marketers, and researchers doing voice-of-customer, social listening, or trend research who want a defensible, recency-weighted brief. It complements the companion blog-researcher skill by adding the practitioner-discourse lens.