Notes Automation Workflows
SkillDocs & knowledgenotes-automation-workflows — Automate bulk Apple Notes workflows: discovery, filtering, and scripted enrichment for links/photos.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Notes Automation Workflows skill
What this skill tells your AI
The instructions your AI receives, as published by atlasomnia/hermes-custom-pack in skills/notes-automation-workflows/SKILL.md and read by ahel’s review.
Use this skill for scripted, repeatable Apple Notes operations that go beyond single-note edits.
When to use
- "Review my notes from last N days" or periodic cleanup/triage tasks.
- Adding structured summaries to notes that contain links.
- Adding summaries for image/photo notes.
- Updating notes programmatically without manual, one-by-one edits.
- Creating a translated or otherwise transformed copy of a rich-text note while preserving its structure.
When not to use
- Single small edits: use the
memo/Notesinteractive flow directly. - Sensitive credential text cleanup: use explicit user confirmation before summarizing.
- Non-Apple notes sources (Obsidian, Notion, etc.)
Prerequisites
- macOS Notes.app
osascriptavailable- User has granted Automation access to Notes.app in System Settings → Privacy & Security → Automation
- Apple Notes has the target notes in a reachable account (usually iCloud)
memo CLI reliability
The memo CLI tool frequently breaks after Homebrew/Python updates due to a stale interpreter path. Diagnose with /opt/homebrew/bin/memo 2>&1 — typical error: bad interpreter: No such file or directory. Fix via brew uninstall memo && brew install antoniorodr/memo/memo. When broken, fall through entirely to AppleScript (see search section below).
Searching notes reliably
Title-only AppleScript query is the fastest, most reliable search method. Use before any other approach:
# Search by title only — instant even with thousands of notes
osascript -e 'tell application "Notes" to get name of every note whose name contains "keyword"'
# Retrieve title + body of a single matched note (use sparingly)
osascript -e 'tell application "Notes" to set target to first note whose name contains "keyword"' \
-e '{name:name of target, body:(body of target)}'
Critical timing constraint: Body access (body of n) iterated across every note in a large account times out at 30s. Title-only search has no this problem. Restrict full-body iteration to single-note retrieval or folder-scoped queries.
Core strategy
For anything beyond one note, use AppleScript first-pass enumeration + AppleScript updates.
Avoid relying on memo index-based addressing for scripts.
For a single note that must be created and shown on screen, follow the quick-create procedure. It covers the memo bad-interpreter fallback, account/folder discovery, HTML-body creation, show/activate, and folder-scoped read-back verification.
Reliable read/write path
Use AppleScript with body (not plain text) and id:
tell application "Notes"
set target to first note whose id is "<note-id>"
set noteBody to (body of target) as string
set noteAttachments to count of attachments of target
end tell
Prefer updating by id for correctness and stability.
14-day sweep pattern
When a user asks for the last X days, start with a filtered enumeration in AppleScript so every candidate is in-bounds before heavy processing.
tell application "Notes"
set cutoff to (current date) - (14 * days)
set out to ""
repeat with a in accounts
set nset to (notes of a whose (modification date >= cutoff))
repeat with n in nset
set out to out & (id of n as string) & "\u001f" & (name of n as string) & "\u001f" & (count of attachments of n as string) & "\u001f" & ((body of n) as string) & "\n"
end repeat
end repeat
return out
end tell
This keeps pagination/sort order issues out of the script and gives a stable candidate list.
Classification rules used in enrichment workflows
After loading candidates, classify each note before writing:
- Link note
- URL regex appears in title or body (
http://orhttps://).
- Photo/attachment note
attachments > 0- stripped text body is minimal/noisy (
len(text) <= ~40) or only boilerplate.
- Regular text note
- No actionable link
- Non-empty readable text and no dedicated attachment-only signals
For link notes, summarize source content and prepend the note body with a single summary block. For photo-only notes, prepend a concise photo summary block (or a clear placeholder if the image is inaccessible).
Idempotent updates
Never duplicate summary blocks. Before writing, check if body already contains your summary marker, e.g.:
if (body of target) does not contain "<b>Summary:</b>" then
set body of target to ("<div><b>Summary:</b> ...</div><div><br></div>") & (body of target)
end if
Content formatting requirements
- Keep all appended summaries as HTML and prepend via
body. - Wrap each inserted piece in
<div>and separate entries with<div><br></div>. - Keep summaries short and task-scoped (2–5 sentences max).
Link-source handling
web_extract coverage varies by site. If a source is unsupported:
- record that as a blocker in the summary
- summarize from cached context only when allowed
- keep a clear label so the next pass can retry when source access improves
Validation checklist
After writes, verify:
- expected notes in scope were touched
- link notes now include one summary block each
- photo/attachment-only notes are marked in a consistent format
- non-link/non-photo notes were not unexpectedly rewritten
Signals
- GitHub stars
- 56
- Forks
- 5
- Last commit
- Aug 2026
Advanced
- Catalog kind
- skill
- Gateway key
notes-automation-workflows- Source
- github.com/atlasomnia/hermes-custom-pack