hierarchical-memory
par balkanblbn
Web & Frontend DevelopmentRésumé
Manage and navigate a multi-layered, branch-based memory system.
Description
Ce que fait ce skill
Organizes agent memory into three layers: Root, Domain, and Project. Each layer is a local markdown file, linked into a navigable hierarchy. A Python script automates creating new branches and inserting the correct links.
Branching memory into focused files means an agent reads only what is needed, avoiding the cost of scanning a single flat memory document on every task.
Quand l'utiliser
- Keeping trading strategy notes separate from coding project details
- Loading only the relevant memory branch for a focused task
- Tracking multiple active projects without losing context between sessions
- Syncing an agent to recent changes via per-domain delta files
- Pruning obsolete context during periodic memory reviews
Exemple de workflow
User asks: Set up a new project memory for my trading bot
- 1
Reads MEMORY.md to locate the relevant domain file
- 2
Runs scripts/add_branch.py to generate memory/projects/trading-bot.md from the boilerplate template
- 3
Populates the file with project status, tech stack, and open to-dos
- 4
Adds a backlink in the trading domain file pointing to the new project file
- 5
Updates MEMORY.md with a link to the new branch
A structured project memory file linked from both the domain and root layers, ready for future recall
Installation
clawhub install balkanblbn/hierarchical-memoryExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.