dory-memory
bởi justinhartbiz
Web & Frontend DevelopmentTóm tắt
File-based memory system for AI agents that forget between sessions.
Mô tả
Kỹ năng này làm gì
File-based memory system for AI agents that lose context between sessions. Implements a structured workspace with state files, boot sequences, and conflict resolution so agents can resume work reliably. Uses exact user quotes to prevent intent drift across context resets.
Most agent memory approaches rely on embeddings or summaries that lose precision; this system stores exact user instructions in plain files, making state auditable and trivially portable.
Khi nào nên dùng
- Resuming a multi-day coding task without re-explaining context
- Tracking blocked items across agent sessions
- Preserving decisions made during long research workflows
- Setting up a persistent workspace for a custom AI assistant
- Preventing repeated mistakes by logging lessons to disk
Quy trình ví dụ
User asks: 'Set up memory for my agent and start tracking my refactor task'
- 1
Copy template files from assets/templates/ to the workspace directory
- 2
Write the user's exact words to state/ACTIVE.md with interpretation and steps
- 3
Run the boot sequence: read HOLD.md, ACTIVE.md, DECISIONS.md, recent-work.md, MEMORY.md
- 4
Output the boot status line showing active task, hold count, and staging count
- 5
Begin work, logging decisions with timestamps to state/DECISIONS.md
Agent workspace initialized with active task tracked; subsequent sessions resume from the same state without re-prompting
Cài đặt
clawhub install justinhartbiz/dory-memoryChạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.