@realpda

hermes-memory-sync extracts daily conversation summaries from agent session logs and writes structured markdown files. covers python script installation, cron scheduling, output format, and supported session file types.

master-research-notetaking systematizes investigation of a thinker via six parallel research vectors (writings, dialogue, external critique, decisions, timeline, applied cases), extracts reusable mental models through triple verification, and writes structured obsidian notes. avoids pedia-copying, prioritizes thinking patterns over biography.

kg-obsidian-mini organizes knowledge graph notes in obsidian using four note types, enforces read-only analysis before execution, and validates against yaml via python scripts. execution checklist workflow separates planning from modification.

deep-token-saver proposes six token reduction layers including output compression (caveman mode), input deduplication, tiered note loading, and persistent memory via remnic/qmd. lacks clear activation triggers and error recovery paths.

Obsidian笔记库按KG笔记法整理——四种笔记类型、链接权限、关系中转、命名规范、新旧分离。自动纠错+建关系+去重+补缺+无关内容分离+分层加载

对用户提出的具体话题/课题/行业进行深度调研，输出结构化报告。结合多智能体并行搜索（借鉴女娲Nuwa）和克制调用原则（借鉴AutoGLM DeepResearch），先展示中间发现再决策。

Agent跨会话记忆——任务前搜agent memory L0→L1→L2分层加载，任务结束LLM提取事实→哈希去重→写入碎片。禁type/[[链接]]/related_fragments