rune
от thebobloblaw
Notes & PKMКратко
Self-improving AI memory system with intelligent context injection and adaptive learning.
Описание
Что делает этот навык
Rune gives an OpenClaw agent persistent memory stored in a local SQLite database. It dynamically selects relevant facts per conversation rather than loading everything at once, and applies forgetting curves so stale information fades over time.
Injects only contextually relevant facts per conversation instead of loading a full static context file, reducing token usage while keeping responses more accurate.
Когда его использовать
- Remembering why a technical decision was made weeks later
- Getting prioritized task suggestions when unsure what to work on next
- Tracking which tools or approaches caused problems in past sessions
- Detecting when a project has stalled and needs attention
- Adapting to whether a session is exploratory brainstorming or focused debugging
Пример сценария
User asks: What should I work on next?
- 1
Runs
rune injectto load relevant facts into session context - 2
Runs
rune next-taskto retrieve priority-scored task recommendations - 3
Checks
rune stuck-projectsto surface any blocked work - 4
Returns the top task with its priority score and relevant memory context
Top recommended task with priority score, relevant project facts, and any identified blockers
Требования
- Ollama installation (optional, required for local relevance scoring and fact extraction)
- OpenAI API key (optional, for cloud-based reasoning features)
- Anthropic API key (optional, for cloud-based reasoning features)
Установка
clawhub install thebobloblaw/runeВыполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.