aria-knowledge или io.github.keshrath/agent-knowledge

ARIA is automated knowledge capture, management, and application for Claude Code. It actively learns during AI-assisted development sessions — insights, decisions, patterns, gotchas — and promotes validated knowledge into an indexed markdown knowledge base that is proactively applied in sessions and on every edit to improve development and accuracy. It runs in Claude Code, fully local. Core feature sets are: Knowledge Lifecycle — automatic capture from conversations and pre-compaction transcripts, staging backlogs, review workflows, promotion to tagged files, staleness detection, and archival. A pipeline from raw signal to validated knowledge. Decision Discipline — hook-enforced impact assessment on every file edit, change framework with scope verification, pre/post edit checks, configurable critical paths, and abbreviated mode for planning docs. Not guidelines — active enforcement. Contextual Retrieval — tag-based indexing with cross-references, on-demand knowledge loading by topic, project tag expansion, en

Cross-session memory for AI agents - knowledge graph, scoring, semantic search