Agent Memory
Persist decisions, preferences, and open threads so the next session does not start blank.
Claude Cursor Any agent memorynotescontinuity
What it is
Agent Memory writes durable notes the model can reload. It is for project conventions, not secrets. Pair it with skills that already tell the agent where facts live.
When to use
Use on long-running repos where the agent keeps re-asking the same product questions.
How to install
Install agent-memory/SKILL.md and point it at a committed notes file, never at .env.
# Claude Code
.claude/skills/agent-memory/SKILL.md
# Grok Build
.grok/skills/agent-memory/SKILL.md
# Codex
.agents/skills/agent-memory/SKILL.mdExample prompt
Remember that we use trailingSlash never and English as the default locale. Write it into agent notes.
Source: Open Agent Skills ecosystem