global-memory
/global-memory
Read and write the user's global memory in <config>/memory/, the cross-project counterpart to per-project auto memory. Use before saving any memory, to decide whether it is global or project-scoped; when a global index line looks relevant and you need the full file; and when the user says "remember this everywhere", "make this a global memory", "move this memory to global", or asks what is in global memory.
- Author
- allir ↗
- Repository
- agent-skills ↗
- Licence
- Apache-2.0
- Last repository change
- 07.11.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 07.11.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
Read and write the user's global memory in <config>/memory/, the cross-project counterpart to per-project auto memory. Use before saving any memory, to decide whether it is global or project-scoped; when a global index line looks relevant and you need the full file; and when the user says "remember this everywhere", "make this a global memory", "move this memory to global", or asks what is in global memory.
Triggers from the description
Original author description
Read and write the user's global memory in <config>/memory/, the cross-project counterpart to per-project auto memory. Use before saving any memory, to decide whether it is global or project-scoped; when a global index line looks relevant and you need the full file; and when the user says "remember this everywhere", "make this a global memory", "move this memory to global", or asks what is in global memory.
How it works
- Setup
- Invoked with no request
- Where a memory goes
- Reading
- Writing
- Moving a memory between stores
SKILL.md preview
---
name: global-memory
description: Read and write the user's global memory in <config>/memory/, the cross-project counterpart to per-project auto memory. Use before saving any memory, to decide whether it is global or project-scoped; when a global index line looks relevant and you need the full file; and when the user says "remember this everywhere", "make this a global memory", "move this memory to global", or asks what is in global memory.
---
# Global Memory
`<config>` is the Claude config directory: `$CLAUDE_CONFIG_DIR` when set, otherwise `~/.claude`. Resolve it once (`echo "${CLAUDE_CONFIG_DIR:-$HOME/.claude}"`) and use the real path everywhere below.
Two memory stores, same file format:
| Store | Path | Loaded |
|---|---|---|
| Global | `<config>/memory/` | Index imported by `<config>/CLAUDE.md`, so every session sees it |
| Project | `<config>/projects/<launch-dir>/memory/` | Only in sessions launched from that directory |
Each store has a `MEMORY.md` index (one line per memory) plus one file per fact. Only the index is in context; read a memory file when its index line looks relevant.
## Setup
The global store only works if every session loads its index. If `<config>/memory/MEMORY.md` is missing, or `<config>/CLAUDE.md` lacks the section below, tell the user and offer to set it up:
1. Create `<config>/memory/MEMORY.md` (empty is fine).
2. Add this section to `<config>/CLAUDE.md`, with `<config>` written out as the real path (imports don't expand variables). The import loads the index into every session; the text tells each session the store exists, since the built-in memory instructions only know the project store:
```markdown
## Global memory
Cross-project memories live in `<config>/memory/`, alongside the per-project auto memory. Facts that hold across repos go global; only single-repo facts go in project memory. Use the `global-memory` skill (`global-memory:global-memory` when installed as a plugin) before reading or writing either.
@<config>/memory/MEMORY.md
```
## Invoked with no request
Show an overview and stop: one bullet per memory in each store (filename, then its index hook), and flag any index line without a file or file without an index line. Don't ask what to do next.
## Where a memory goes
**Global** when the fact holds regardless of which repo the session started in: how the user wants things done, their tools and machine setup, accounts and environments, workflows that span repos, knowledge about the team or organization.
**Project** only when the fact is about one repo and means nothing elsewhere: a test quirk, a build gotcha, a file layout.
When unsure, go global. A project memory is invisible to every other session.
## Reading
1. Scan the global index (already in context) and the project index.
2. Read the full file for any line that looks relevant to the task.
3. Memories are point-in-time. Verify a file path, flag or command still exists before relying on it.
## Writing
1. Check both indexes for a memory that already covers it, and update that file instead of adding a duplicate.
2. Write one fact per file, named `<type>_<topic>_<fact>.md` (e.g. `feedback_git_annotated_tags.md`). Reuse a topic already in the store when one fits, so related facts sort together, and keep the index sorted by filename:
```markdown
---
name: <short-kebab-case-slug>
description: <one-line summary, used to decide relevance>
metadata:
type: user | feedback | project | reference
modified: <YYYY-MM-DD>
---
<the fact; for feedback/project, follow with **Why:** and **How to apply:** lines>
```
3. Link related memories with `[[slug]]`. Slugs resolve across both stores.
4. Add `- [Title](file.md) — hook` to that store's `MEMORY.md`. Keep the line short, since the global index loads in every session.
5. Delete memories that turn out to be wrong, and remove their index line.
## Moving a memory between stores
`mv` the file, then move its index line from one `MEMORY.md` to the other. Slugs don't change, so links keep working.
Versions
The database has no other skills named global-memory.
Database snapshot: 13.11.2026 · package version 2051a495f2a3 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100cortex-remember-global
Store a global memory that is visible across all projects. Use when the user shares architecture rules, coding conventions, infrastructure facts, security policies, team agreements, or any knowledge that applies beyond a single project. Triggers on 'remember this everywhere',… - 90memory-save
When the user wants to save or remember something, use this skill to choose between add_memory and add_wiki and store it correctly. - 100knowledge
Capture durable project or global knowledge into TerMinal's Knowledge Base. Use on /knowledge, when saving links/media/snippets/references, or when the user asks to remember useful context visually. - 90global-audit
Audits and optimizes the five global Claude layers (User Preferences, Styles, Global Memory, Skills, MCP Connectors) using the Global Layer Scorecard (six dimensions, anchored 1-5 rubrics). Detects eight cross-layer failure modes and produces evolutionary recommendations… - 100memory-neat-freak-desktop
Project-memory hygiene for desktop Agent workspaces with optional explicit global memory. Use whenever the user asks to organize, migrate, distill, capture, clean up, or retire .omx, .ccg, doc(s)/superpowers, task_plan.md, findings.md, progress.md, AGENTS.md, README.md,… - 90rootnode-global-audit
Audits and optimizes the five global Claude layers (User Preferences, Styles, Global Memory, Skills, MCP Connectors) using the Global Layer Scorecard (six dimensions, anchored 1-5 rubrics). Detects eight cross-layer failure modes and produces evolutionary recommendations…