Notes Archive Skill Skill

Notes Archive Skill

Archive Structure

Location: $NOTES_PATH (defaults to ~/Dropbox/Notes if not set)

Directory Hierarchy: YYYY/MM/YYYYMMDD_ID[_slug].md

Examples:

2026/01/20260106_8823.md - no slug
2026/01/20260102_8814_todo.md - with slug
2024/12/20241203_6973_disable-letter_opener.md - descriptive slug

ID Tracking: Sequential IDs stored in $NOTES_PATH/id.json

UID (Unique Identifier): YYYYMMDD_ID uniquely identifies each note

File Format

YAML Frontmatter (Optional):

---
title: Note title
tags: [tag1, tag2, tag3]
slug: short-name
description: Brief description
---

Frontmatter rules:

Only include fields that are explicitly requested or clearly applicable
Do NOT add a date field — date is already encoded in the filename
Add description when there is clear context behind the note (e.g., responding to a message, capturing a decision, summarizing a conversation) — use it to record the context/intention, not a summary of the body content
If there is no clear context/intention behind the new note, prefer not to include description field (avoid generic/non-informative descriptions)

Task Syntax (in *_todo.md notes):

[+] - Completed
[>] - In progress
[ ] - Pending
[daily] - Tag for daily recurring tasks

Agent Scripts

Location: bin/ subdirectory within this skill (~/.claude/skills/notes-archive/bin/)

Usage: When this skill is invoked, reference scripts as: ~/.claude/skills/notes-archive/bin/SCRIPT_NAME

create-note `[slug]`

Creates note with auto-incremented ID, updates id.json

Input: Content from stdin
Output: Created file path to stdout
Example: echo "# My Note" | ~/.claude/skills/notes-archive/bin/create-note my-slug

create-todo `[--force]`

Creates today's todo note by carrying over incomplete tasks from previous todo

Finds most recent todo note and extracts pending ([ ]) and in-progress ([>]) tasks
Daily tasks (marked with [daily]) are always included
Marks copied pending tasks as in-progress in previous todo
Creates new todo with slug todo and auto-incremented ID
If today's todo already exists, returns existing path (use --force to regenerate)
Output: Created/existing file path to stdout
Example: ~/.claude/skills/notes-archive/bin/create-todo

latest-note-path

Outputs path to most recent note (skips 99999999_* pinned notes)

latest-todo-path

Outputs path to most recent *_todo.md note

latest-backlog-path

Outputs path to most recent *_backlog.md note

latest-weekly-path

Outputs path to most recent *_weekly.md note

dump-chat `<session-id-or-jsonl-path> <output-notes-path>`

Exports a Claude Code chat session to a Markdown notes file.

Reads the session's .jsonl file from ~/.claude/projects/
Pulls title/summary from sessions-index.json, falls back to first user message
Preserves: session ID, model name, Claude Code version, project path, git branch, creation timestamp, per-message timestamps
Skips tool results, thinking blocks, and sidechain messages — primary dialog only
Output file is overwritten if it already exists (idempotent for updates)
Templates for heading and messages are defined at the top of the script for easy customisation
Example: ~/.claude/skills/notes-archive/bin/dump-chat ade243b9-7ea3-4fa0-8277 ~/Dropbox/Notes/2026/02/20260218_9004_chat-dump.md

Error handling: All scripts check for $NOTES_PATH and exit with error to stderr if missing or invalid.

Pinned Notes: Files starting with 99999999_* are excluded from latest-* searches.

Common Search Patterns

Find by ID or slug:

find . -name "*_8823*.md"  # by ID
find . -name "*_todo.md"    # by slug

Find by date:

ls 2026/01/
ls 2026/01/20260106_*.md

Search content:

grep -r "search-term" 2026/
grep -r "^tags:" --include="*.md"

Active tasks:

grep -r "^\[>\]" --include="*_todo.md"

Tag analysis:

# Tag frequency
grep -rh "^tags:" --include="*.md" | sed 's/tags: \[\(.*\)\]/\1/' | tr ',' '\n' | sed 's/^ *//;s/ *$//' | sort | uniq -c | sort -rn

# Common slugs
find . -name "*_*.md" | sed 's/.*_\([^.]*\)\.md/\1/' | grep -vE '^\d+$' | sort | uniq -c | sort -rn | head -10

Usage Guidelines

Create notes: Use bin/create-note or bin/create-todo scripts
Access notes: Use bin/latest-*-path scripts to get file paths, then Read tool to view
Search: Use Grep/Glob tools or bash commands from "Common Search Patterns" section
Editing: Preserve YAML frontmatter structure when modifying notes
References: Use UIDs (YYYYMMDD_ID) for cross-referencing notes

Chat History Dumps

Use bin/dump-chat to save the current (or any past) Claude Code chat session as a note.

Finding the current session ID

The current session's .jsonl file is the most recently modified file in: ~/.claude/projects/<project-dir>/

Use: ls -lt ~/.claude/projects/<project-dir>/*.jsonl | head -1

The filename (without .jsonl) is the session ID.

Workflow

"Dump this chat" (first time or explicitly new):

Find current session ID from most recent .jsonl
Create new note: ~/.claude/skills/notes-archive/bin/create-note chat-dump (pipe empty content or use touch)
Run: ~/.claude/skills/notes-archive/bin/dump-chat <session-id> <new-note-path>
Report the note path

"Dump this chat" (repeated request, no "new file" / "copy" mentioned):

Find the most recently modified *_chat-dump.md note (Glob: $NOTES_PATH/**/*_chat-dump.md, pick newest mtime)
Run: ~/.claude/skills/notes-archive/bin/dump-chat <session-id> <existing-note-path> (overwrites in place)
Report the note path

"Save a copy" / "save to a new file": Always create a new note regardless.

`/dump` slash command

When the user types /dump, treat it as "dump this chat to a note" using the workflow above (update existing chat-dump note if one exists for today, otherwise create new).

Agent Skills: Notes Archive Skill

Install this agent skill to your local

Skill Files