memory-to-vault

name	memory-to-vault
version	2.0
category	utilities
execution_speed	slow
token_efficiency	low
triggers	["archive session knowledge","convert memory to vault","structured vault outline","organize accumulated knowledge","memory outline"]
cache_key	memory-to-vault-2.0
dependencies	[]
description	Convert Claude's persisted memory into a structured high-level outline and save it to the project Obsidian vault. Use to archive and organize accumulated session knowledge as a navigable vault note.
disable-model-invocation	true

Memory to Vault

Reads Claude's persisted memory for the current project and generates a categorized high-level outline, saved as a vault note.

Resilience (checkpoint + resume)

Follow the global Long-Workflow Resilience rule (CLAUDE.md): on an API error (500, 529 overload, usage limit) while reading memory or writing the outline, stop cleanly and tell the user to re-run rather than half-writing the vault note. The generated note is a single artifact -- no per-item checkpoint needed.

What This Generates

A Memory Outline.md vault note containing:

• User Profile -- who the user is, role, expertise, working style
• Behavioral Guidelines -- what to avoid, what to repeat, confirmed approaches
• Project Context -- active initiatives, goals, architectural decisions
• References -- external system pointers, tools, URLs

No conversation history. No session logs. No raw user responses.

Process

1. Find memory directory

Run the script to locate and read all memory files:

# Windows
& ".claude/skills/memory-to-vault/scripts/memory-to-outline.ps1"

# Unix -- use equivalent shell logic
hash=$(pwd | sed 's|/|-|g; s|^-||')
memdir="$HOME/.claude/projects/$hash/memory"

The script prints all memory file contents to stdout, organized by type.

2. Read memory index

Read MEMORY.md first -- it is the index. Then read each linked file.

Memory types and what they tell you:

• user_*.md -- user's role, expertise, preferences
• feedback_*.md -- behavioral rules (lead with rule, then Why: and How to apply: lines)
• project_*.md -- active work, goals, deadlines, decisions
• reference_*.md -- where to find things: Linear, Slack, Grafana, etc.

3. Generate outline

Organize content into four sections. For each memory entry write:

• One-line summary of the rule/fact
• Sub-bullets for key details
• Omit: conversation snippets, session-specific state, anything ephemeral

4. Determine vault path

Find vault in this order:

1. docs/Obsidian Vault/[Project Name]/ in current repo
2. /mnt/d/Obsidian Vault/AI Research/ (legacy global path)
3. Ask the user if neither found

5. Write Memory Outline.md

Save to vault with this format:

# Memory Outline

Generated: [YYYY-MM-DD]
Source: [memory directory path]

## User Profile
[bullet per user memory]

## Behavioral Guidelines
[bullet per feedback memory -- rule + why + how to apply]

## Project Context
[bullet per project memory -- fact + why + how to apply]

## References
[bullet per reference memory]

## Related
[[Claude-ITect-Skill Update Index]]

6. Update vault index

Find the main index note in the vault. Add [[Memory Outline]] to its links if not already present.

Notes

• Memory is a snapshot -- it reflects what was known at generation time
• Re-run to refresh after long sessions accumulate new memories
• The outline is a read-only artifact -- edit source memory files directly, not the outline
• If memory directory not found, report the expected path and ask user to verify

Integration with Other Skills

• /obsidian-vault -- Creates and manages the vault this skill writes to.
• /llm-wiki -- The LLM wiki pattern. Memory Outline can seed a new wiki's index.
• /grill-with-docs -- Creates CONTEXT.md. Complements memory outline for project context.