brain-new

Step 1: Accept Input

If the user provided text alongside the command, use that as the input. If no text was provided, prompt: "What do you want to capture? Paste or type anything -- tasks, notes, ideas, logs, all mixed together is fine."

Step 2: Load Vault Context

Silently read:

• Machine/Memory/context-cache.md -- to understand current projects and context
• Machine/Memory/entities.md -- to identify known entities for wikilinks
• Machine/Rules/active-rules.md -- for any routing rules

Step 3: Parse and Classify

Analyze the raw input and break it into distinct items. For each item, classify its type:

Type	Indicators	Destination
Task	Action verbs, "need to", "TODO", "should", deadlines	Today's daily note (Human/Daily/{today}.md) under ## Tasks
Note	Observations, thoughts, facts, context	Human/ -- new note file or append to existing relevant note
Project idea	New project concept, feature idea, "we should build"	Human/Projects/ as a new project note
Decision	"Decided to", "going with", choice between options	Machine/Memory/decisions.md
Entity info	Facts about a person, tool, project, org	Machine/Memory/entities.md
Meeting/event note	"Met with", "call about", timestamps	Human/Daily/{today}.md under ## Notes
Log/status update	Progress report, "done with", "finished"	Today's daily note under ## Notes
Idea/someday	"Maybe", "someday", "it would be cool if"	Human/Projects/ideas.md or Human/Archive/someday.md
Correction	"Don't do X", "always do Y", behavioral instruction	Machine/Memory/corrections.md
Reference/link	URLs, book titles, article references	Human/ as a reference note with source metadata

Step 4: Route Each Item

For each classified item:

1. Determine the destination file based on type and context.
2. Check if the destination file exists. If yes, append. If no, create with proper frontmatter.
3. Apply formatting:
   • YAML frontmatter on new files: date, tags, type
   • [[wikilinks]] for any references to known entities or existing notes
   • Proper markdown structure (headers, lists, checkboxes for tasks)
   • Tags based on content (e.g., #project/name, #meeting, #idea)
4. Write the item to the destination.

Step 5: Update Daily Note Log

Append a processing log to today's daily note under a ## Captured section:

## Captured
<!-- Auto-logged by new skill at HH:MM -->
- [task] "Fix the login bug" --> added to Tasks
- [note] "API rate limits are 100/min" --> created [[API Rate Limits]]
- [idea] "Build a CLI dashboard" --> added to [[Ideas]]
- [decision] "Using PostgreSQL over SQLite" --> logged to decisions.md

Step 6: Confirm to User

Present a summary of what was processed:

Processed {N} items from your brain dump:

- 2 tasks --> added to today's daily note
- 1 project idea --> created [[Project Name]]
- 1 decision --> logged to decisions.md
- 1 entity update --> updated [[Entity Name]] in entities.md

Anything I miscategorized or missed?

<success_criteria>

• All items from the input are classified and routed
• Each item lands in the correct vault location
• Proper frontmatter, tags, and wikilinks are applied
• Today's daily note has a log of what was processed
• User receives a clear summary of routing decisions
• No input is silently dropped </success_criteria>