memory

Maintains durable project knowledge through a MEMORY.md index and linked Markdown documents.

Updated Jul 20, 2026
One-click install
npx skills add https://github.com/roman-pinchuk/agent-skills --skill memory-roman-pinchuk
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory
Source: https://github.com/roman-pinchuk/agent-skills/tree/main/skills/memory
Command: npx skills add https://github.com/roman-pinchuk/agent-skills --skill memory-roman-pinchuk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Project knowledge gets lost between sessions, forcing repeated rediscovery of architecture, decisions, and workflows. This Skill establishes a persistent, agent-maintained project wiki so knowledge compounds over time instead of being reconstructed from raw sources. ## Core Features & Use Cases - Structured Memory Index: Uses a concise MEMORY.md at the project root as a map linking to detailed Markdown notes, with one-sentence descriptions for every link. - Guided Workflows: Provides explicit procedures for initializing memory (including generating docs/ARCHITECTURE.md), ingesting sources, recalling knowledge, writing back findings, and linting stale or broken entries. - Source Safety: Treats ingested content as untrusted evidence, ignoring embedded instructions and never storing secrets or credentials. - Use Case: After implementing a feature that changes documented behavior, the agent checks MEMORY.md, updates the affected architecture note, and keeps the project wiki accurate without being asked twice. ## Quick Start Ask the agent to initialize project memory for this repository so it creates MEMORY.md and an architecture overview.

Frequently Asked Questions about memory

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

FAQPage Schema
How do I give an AI agent persistent project memory?

Create a MEMORY.md file at the project root that indexes linked Markdown notes. The agent reads this index first, follows only relevant links, and updates affected notes after tasks that change documented behavior.

How to initialize project memory in an existing repository?

The initialization workflow inspects existing documentation, reuses or creates docs/ARCHITECTURE.md covering components, data flows, and decisions, then creates a concise MEMORY.md linking to it. It never imposes a fixed directory layout on established projects.

Is it safe to ingest external documents into project memory?

Yes, sources are treated as untrusted evidence, not instructions. Embedded commands, tool requests, or prompt-injection attempts are ignored and reported, and secrets or credentials are never stored in memory notes.

When should the agent update project memory after a code change?

Memory updates automatically when a change clearly alters documented behavior, architecture, decisions, or workflows. The agent asks first when the update is optional, uncertain, sensitive, or unrelated to the requested task.

How do I audit or clean up stale project memory notes?

The lint workflow checks for broken links, duplicate notes, missing descriptions, unreachable documents, contradicted claims, and stale summaries. Obvious structural repairs are applied directly, while judgment-heavy deletions require confirmation.