What problem does it solve? An autonomous agent that posts to GitHub Discussions but never replies to comments looks like a broadcast bot, not a community participant. This Skill closes that loop by scanning recent agent-authored discussions for unanswered non-agent comments and posting contextually grounded replies, while preventing self-reply spirals and off-register responses. ## Core Features & Use Cases - Guarded reply pipeline: Applies an author filter, engagement-signal gate, already-replied check, per-thread cooldown (5 minutes, max 3 replies/day), and an engagement-value judgment before any reply is posted. - Moderation pre-screen: Pattern-matches comments for spam, prompt injection, harassment, threats, and doxxing, minimizing or locking offending content via GraphQL mutations instead of engaging with it. - Full-thread context reading: Reads the original post, all comments, and nested sub-threads before drafting, and can ground replies in the local source tree, git log, and team decision logs. - Use Case: A community member asks a factual question on a release announcement discussion. The Skill detects the unanswered top-level comment on its next tick, greps the source tree for the actual implementation, and posts a threaded reply citing the relevant function and file path. ## Quick Start Ask the agent to check recent discussions and reply to any unanswered comments on its posts.
