AlekseiUL/agentforge-openclaw

Here's a technical summary for the developer directory: Implements a 9-step pipeline with tiered memory architecture: contextual (session), file-based (lessons/patterns/logs), vector search (past conversations), and identity layers (role/personality/user profile). The auto-handoff system periodically snapshots conversation state to markdown, enabling 95% context recovery after resets versus the typical 30-50% loss. Provides structured templates for three agent archetypes (Full/Specialized/Mask) and four skill types (Workflow/Role/Data-driven/Hybrid), plus pattern-driven self-improvement that converts repeated mistakes into permanent rules. --- **Word count: 71 | Technical specificity: Memory architecture, context recovery mechanism, template-based archetypes**