skills.rest
t0lab
t0labOFFICIAL
Explore all skills by this author →
WebsiteGitHub

memory

Persist and recall long-term conversational and project memory across sessions.

2|Updated Apr 10, 2026
Search Tags:#memory#knowledge-management#project-memory#long-term-memory#user-preferences#hot-cache#deep-store
One-click install
npx skills add https://github.com/t0lab/harness-kit --skill memory-t0lab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: memory
Source: https://github.com/t0lab/harness-kit/tree/main/.agents/skills/memory
Command: npx skills add https://github.com/t0lab/harness-kit --skill memory-t0lab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Projects and conversational agents lose important context between sessions: user preferences, corrections, design decisions, and external references are frequently forgotten or scattered across files and chat history. This Skill provides a repeatable, safe process to capture, store, and retrieve those facts so future sessions can act consistently and avoid re-asking or re-evaluating settled decisions.

Core Features & Use Cases

  • Typed memory model: explicit user, feedback, project, and reference memory types with recommended storage locations for local and repo-scoped data.
  • Hot cache vs deep store: index-only hot-cache files for context-window efficiency and separate deep-store markdown files with frontmatter for atomic, auditable records.
  • Operational workflows: clear write and recall flows, promotion/demotion rules to manage cache bloat, confidence scoring for feedback memories, and end-of-session compaction guidance.
  • Use case: remember a user's tool preference and apply it across sessions, record why a project architecture decision was made, or store pointers to external systems for quick retrieval.

Quick Start

Tell the assistant to remember a preference or decision, for example: Remember that I prefer dark mode and store it as a user memory so future sessions honor the preference.

Frequently Asked Questions about memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I persist user preferences and project decisions across AI chat sessions?▼

Long-term project and user memory persists conversational context across sessions by storing preferences, feedback, and decisions in local memory or repository .claude/memory files. This maintains continuity so future sessions act consistently without re-asking settled questions.

What's the best way to manage context window bloat when storing long-term agent memory?▼

Managing context window bloat uses an index-only hot cache for active retrieval alongside separate deep-store markdown files with strict frontmatter. Promotion and demotion rules move records between tiers, keeping the context window efficient while retaining auditable details.

How does feedback memory handle conflicting user corrections over multiple sessions?▼

Feedback memory handles conflicting corrections through confidence scoring applied to user feedback memories. This scoring system weights recent and repeated corrections, ensuring the agent prioritizes the most reliable preferences when recalling context in future sessions.

Can I store external system references and project architecture decisions in repository files?▼

You can store external system references and project architecture decisions in repository-scoped .claude/memory files. The typed memory model explicitly supports project and reference memory types, writing atomic records with frontmatter to maintain shared context across teams.

How do I update CLAUDE.md memory indexes without creating duplicate entries?▼

Updating CLAUDE.md and INDEX.md uses idempotent update workflows to prevent duplicate entries. This ensures repeated memory write operations produce identical results, maintaining a clean index of hot-cache pointers without duplicating existing records.

Related Skills

Explore More Software Engineering Skills→
swannysecswannysec
3

memory-init

Create ConKeeper memory directories and starter documents in a git project.

Community
Intermediate
richfremrichfrem
5

memory-management

Orchestrates a six-tier memory system across agent sessions via boot, learning, and seal phases.

Community
Advanced
yusaduymazyusaduymaz

gsd-discuss-phase

Ask targeted questions and consolidate prior decisions into planning-ready context.

Community
Advanced
Sir-chawakornSir-chawakorn
33

memory-system

Manage AI context and memory persistence using structured markdown files.

Community
Intermediate