auto-process-a

Action skill — Inbox processing workflow: read intent, determine destination, confirm, file, update living docs.

Automated Inbox Processing

Process inbox items: read intent, classify, file, rename, connect, and update living docs. Designed for the librarian agent but usable standalone.

Input

• Specific file(s): Process only the named files in 0-Inbox/
• No argument: Process all files in 0-Inbox/

Process Per Item

1. Read and Understand Intent

Read the note. Check frontmatter for intent signals left by the writer or thinker:

Field	Meaning
filing-hint	Suggested destination (area name, project name, "resource", "unsure")
context	Freeform context — what prompted this, what it relates to
source: thinker	Came from a synthesis session — handle with care
type	dump, note, clip, edit
tags	Knowledge role: seed, thread, tool, question, insight

If no intent signals, classify from content.

2. Determine Destination

Filing cascade (first match wins):

1. Project? → Does this help an active project? Check project briefs for goal alignment.
2. Area? → Does this relate to an ongoing area? Match against area scopes.
3. Resource? → Is this reference material? Route to correct sub-folder (clips/short-form/long-form/books/videos).
4. None? → Archive or flag for user.

When filing-hint exists, treat as strong signal but verify.

3. Confirm with Semantic Search

qmd vsearch "key content from the note" --files

Auto-file when:

• Semantic search confirms (3+ top results in same location)
• Content clearly matches one area/project's scope
• filing-hint aligns with search results

Ask the user when:

• Top matches span multiple areas
• Content is vague or multi-topic
• filing-hint: unsure or missing with no clear match

4. Archive Original

Preserve the raw capture before modifying. Rename to YYYY-MM-DD_descriptive-slug.md but keep frontmatter and content as-is — naming is the one thing we always enforce, even on raw archives:

mkdir -p "04-archive/inbox"
cp "0-Inbox/original-name.md" "04-archive/inbox/YYYY-MM-DD_slug.md"

5. Fix Frontmatter

• Set type: note if missing (keep original type if appropriate)
• Set created: YYYY-MM-DD using today's date if missing
• Validate tags against the 5-tag vocabulary (#seed, #thread, #tool, #question, #insight)
• Set appropriate status (draft, refining, or omit for stable)
• Remove intent fields: Strip filing-hint, context, source — they served their purpose

6. Rename and Move

Rename to YYYY-MM-DD_descriptive-slug.md, then move to destination:

mv "0-Inbox/original-name.md" "2-Areas/area-name/YYYY-MM-DD_slug.md"

7. Update Living Docs

Read the parent MOC or project brief. Add the note with annotation:

- [[YYYY-MM-DD_slug]] — brief description of what the note covers

For thinker synthesis (source: thinker):

• Check if the synthesis updates the MOC's thesis paragraph
• Check if it resolves any Open Questions
• Update source notes' ## Related sections

8. Build Connections

Find related notes via semantic search and structural analysis. Add ## Related section if strong connections exist. Only add connections that are clear and useful.

Special: Edit Notes

For notes with type: edit:

1. Read the targets array from frontmatter
2. Read each target file
3. Apply the described edits
4. Archive the edit note

Processing Report

After processing, show summary:

## Inbox Processing Report

### Filed
- "Raw thoughts on agents" → 2-Areas/ai-dev-ecosystem/2025-02-07_agent-thoughts.md
- "Market data notes" → 1-Projects/polymarket-framework/2025-02-07_market-data.md

### Living Docs Updated
- 00_ai-dev-ecosystem.md — added link to agent-thoughts
- 00_polymarket-framework.md — added link to market-data

### Connections Found
- agent-thoughts ↔ agentic-primitives (shared concepts)

### Needs Attention
- "Ambiguous note" — couldn't determine destination, left in inbox

After Batch Processing

Reindex semantic search:

qmd update