remember
Explicitly save important knowledge to auto-memory with timestamp and context. Use when a discovery is too important to rely on auto-capture.
Security Assessment
About remember
This skill provides an explicit '/si:remember' command that saves an important fact, convention, or hard-won insight to a project-scoped auto-memory file so Claude will see it at the start of every future session. It addresses the problem of losing valuable context between sessions: rather than relying on Claude to automatically notice and retain a discovery, the user can deliberately pin it. It is meant for knowledge important enough that automatic capture is too risky to depend on.
The workflow parses the user's input into a concrete fact, optional rationale, and scope; checks the project's MEMORY.md for duplicate entries via a keyword grep and offers to update rather than duplicate; then appends a concise one-line entry. It warns when MEMORY.md approaches a 200-line budget, suggests promoting rule-like statements to CLAUDE.md via a companion '/si:promote' command, and confirms what was saved. The documentation emphasizes concise, actionable entries (concrete commands and values over vague concepts) and gives examples such as build/test commands, project conventions, tool gotchas, and architecture decisions.
It targets developers using Claude as a coding assistant across recurring sessions on a project, especially those who want a lightweight, reviewable memory layer. Notably, the skill's own guidance explicitly instructs users never to store credentials, tokens, or secrets in memory files, and to route enforced rules and cross-project knowledge to CLAUDE.md instead. Use cases include capturing debugging insights, CI requirements, and personal preferences that should persist.
FAQ
What does /si:remember actually do?
It appends a concise one-line entry to the project's auto-memory file (MEMORY.md), which Claude reads at the start of every session in that project, so the fact persists across sessions.
How does it avoid duplicate memories?
Before writing, it runs a keyword grep against MEMORY.md. If a similar entry already exists, it shows it and asks whether to update the existing entry or add a new one.
Is there a limit to how much I can store?
The memory file targets a roughly 200-line budget. When it passes about 180 lines, the skill warns you and suggests running a memory review to free space.
When should I use this instead of CLAUDE.md?
Use /si:remember for project facts and notes. For enforced rules (imperatives, always/never conventions) or cross-project knowledge, the skill recommends promoting to CLAUDE.md, which has higher priority.
Can I store secrets or credentials in memory?
No. The skill explicitly warns never to store credentials, tokens, or secrets in memory files; those should be kept out of persisted context entirely.
Install remember
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
alirezarezvani/claude-skills