What problem does it solve? Keeping an AI agent harness current requires manually tracking dozens of changelogs, blogs, GitHub releases, and YouTube channels, then figuring out which updates actually apply to your setup. This Skill automates that sweep and filters every finding against your system's verified current state so already-implemented or previously rejected ideas are never re-recommended. ## Core Features & Use Cases - Multi-source monitoring: Checks 30+ Anthropic sources (blogs, changelogs, GitHub repos, docs) plus RSS feeds, YouTube channels, and GitHub trending via a Bun script with diff-based state tracking. - Grounded recommendations: Every suggestion carries a Prior Status tag (NEW/PARTIAL/DISCUSSED/REJECTED) backed by file:line evidence gathered during the run, with quoted techniques mapped to specific files. - Internal reflection mining: Mines the system's own algorithm reflection logs to surface recurring weaknesses, cross-referencing internal pain with external techniques. - Use Case: Run a weekly upgrade check that scans Anthropic's claude-code changelog and top practitioner blogs, then receive a tiered report (CRITICAL/HIGH/MEDIUM/LOW) of concrete techniques with exact target files and implementation snippets. ## Quick Start Ask the agent to check for upgrades and report new techniques from Anthropic sources relevant to the current system.