neurongraph/skills/obsidian-daily-process

Instructions

This skill orchestrates the full Obsidian vault processing pipeline: audio transcription, transcript classification and filing, and triggering of downstream vault update processes.

Environment Setup

This workflow requires environment variables from .env: ASR_CLI, MODEL_PATH, AUDIO_TEMP_DIR, IDEAS_INBOX.

If variables are undefined, prefix bash commands with:

set -a; source .env; set +a &&

Important: The current working directory should be the Obsidian vault root (the directory containing .obsidian/).

Step 1: Invoke the process-audio Skill

Run the process-audio skill to transcribe all audio files in $AUDIO_TEMP_DIR/inputs.

It will check prerequisites, convert and transcribe all audio, and save transcripts to $AUDIO_TEMP_DIR/transcripts/. Wait for it to complete before continuing. If it stops due to missing prerequisites or no files found, stop here too.

Step 2: Verify Obsidian Vault Root

This workflow assumes the current working directory is the Obsidian vault root (the directory containing .obsidian/). Verify:

[ -d ".obsidian" ] && echo "✓ Vault root confirmed"

If .obsidian/ is not in the current directory, the user should:

1. Check if OBSIDIAN_VAULT environment variable is set (it will be loaded in the Environment Setup step)
2. If set, change to that directory: cd "$OBSIDIAN_VAULT"
3. If not set, ask the user to provide the vault path or cd into it before running this workflow

All subsequent file paths are relative to the vault root (e.g., KB_2/00. Inbox/02. Tasks/todo.md).

Step 3: Read Daily Notes Configuration

Read <vault>/.obsidian/daily-notes.json to find:

1. The naming convention for daily notes (default: YYYY-MM-DD.md)
2. The folder where daily notes are stored, relative to the vault root (default: vault root)
3. The template file path, if specified

If daily-notes.json does not exist, use the defaults and continue.

If a template file is specified, read it now — you'll need it when creating a new daily note. Note that the path can contain spaces and You might have to add a .md at the end of the path to find the file.

Step 4: Classify and File Each Transcript

For each .txt file in $AUDIO_TEMP_DIR/transcripts/, first classify it, then file it accordingly.

Classification

Read the transcript and determine its type using these rules, applied in order:

1. Daily Note — The transcript opens with a date reference. Check the first ~20 characters for day names (Monday, Tuesday, Wednesday, Thursday, Friday, Saturday, Sunday, yesterday, today, tomorrow) or date patterns (May, Jan, Feb, etc., or YYYY-MM-DD format). The date in the opening words determines which daily note it belongs to.

   if echo "$transcript" | grep -qi "^[^a-z]*\(monday\|tuesday\|wednesday\|thursday\|friday\|saturday\|sunday\|yesterday\|today\|tomorrow\|january\|february\|march\|april\|may\|june\|july\|august\|september\|october\|november\|december\)"; then
     # Daily Note
   

2. Idea — The transcript contains the word "idea" (case-insensitive) in the first 5 words. Simple check without complex patterns:

   if echo "$transcript" | head -c 100 | grep -qi "idea"; then
     # Idea
   

3. Todo — Everything else. Task-oriented language ("I need to", "remind me", "follow up", "don't forget", "action item") with no date reference at the start and no mention of "idea" signals a todo or task capture.

A transcript matches the first rule that fits — so a note that opens with a date is always a Daily Note even if it also mentions "idea" or tasks.

Filing by Type

Daily Note:

• Derive the target date from the opening words. If relative ("yesterday", "Monday"), resolve it to an absolute date.
• Build the note path from the vault folder, naming convention, and target date.
• If the note already exists, append at the end. If it doesn't exist, create it from the template (substituting the correct date wherever the template uses a date placeholder), then append.
• Format the appended block as:

## Voice Memo

<transcript content>

If multiple transcripts land on the same date, each gets its own ## Voice Memo section.

Idea:

• Check the .env file for the IDEAS_INBOX variable. This points to a folder or directory where idea files should be stored (e.g., KB_2/01. Ideas/).
• If IDEAS_INBOX is set, create a new .md file in that directory with a filename based on the first few words of the idea (e.g., New product feature.md).
• Add the #ideas tag at the end of the file and include the full transcript content.
• If IDEAS_INBOX is not set, ask the user where ideas should be filed in their vault and confirm before filing.

Todo:

• Use obsidian-todotxt Workflow E to locate the todo file (session cache first, then data.json).
• Use obsidian-todotxt Workflow A to append the transcript as a new task at the top of the inbox.

Tell the user the type assigned to each transcript and where it was filed.

Step 5: Wiki Update Pipeline

(Placeholder — not yet implemented.)

This step will trigger an Obsidian wiki update process — scanning newly filed content for entities, concepts, or references that should be reflected in the vault's wiki or reference notes. Leave this step as a no-op for now and note it to the user as "coming soon."

Step 6: ArtMind Knowledge Graph Update

(Placeholder — not yet implemented.)

This step will trigger an artmind knowledge graph update — propagating newly captured ideas, tasks, and notes into the knowledge graph for cross-linking and insight generation. Leave this step as a no-op for now and note it to the user as "coming soon."

Step 7: Full Cleanup

List all temporary files that were created during processing, grouped by directory:

• Original audio files in $AUDIO_TEMP_DIR/inputs/
• Converted WAV files in $AUDIO_TEMP_DIR/outputs/
• Transcript text files in $AUDIO_TEMP_DIR/transcripts/

Ask the user whether to delete them. Delete only on explicit confirmation, and only after all vault notes have been successfully written.

Step 8: Todo Review and Action

Use obsidian-todotxt Workflow F to fetch and display the top 10 uncompleted todos ranked by composite urgency score. This surfaces the highest-priority work — including any tasks just added from this session's transcripts.

Then use obsidian-todotxt Workflow G to ask the user if they want to action one of the todos. If yes, invoke obsidian-todo-action for the selected task.