@dylanbaker24

linkedin-monitor polls linkedin inboxes hourly with configurable autonomy levels (monitor-only to full autonomous), drafts replies in your voice, and manages state deterministically to prevent duplicate alerts.

memory-hygiene audits and cleans lancedb vector storage to reduce token waste from irrelevant auto-recalls. provides commands to wipe corrupted memory, reseed from canonical facts, and automate monthly maintenance cycles.

last30days-lite searches reddit, x/twitter, and the web for discussions from the past month, then synthesizes findings into patterns, mistakes, techniques, and executable prompts. no api keys required.

linkedin-inbox monitors unread messages via ui automation, classifies intent, drafts replies matching user communication style, and routes to discord for approval before sending. scheduled scans respect active hours and quiet periods.