name: inbox-processor description: This skill should be used when processing files in the inbox/ directory, organizing unstructured input, classifying content, or extracting actionable information. Triggered by requests like "process inbox", "organize inbox files", "classify this memo", or "inbox を整理".
Inbox Processor
Overview
Processes and organizes unstructured input files from the inbox/ directory. Classifies content, extracts actionable information, and suggests appropriate destinations for integration into the research workflow.
Core Capabilities
1. File Classification
Classify inbox files into categories based on content.
Categories:
- Meeting Notes: Discussion records, decisions, action items
- Protocols: Experimental procedures, SOPs
- Ideas: Research ideas, hypotheses, brainstorming
- Data: Raw data files, datasets
- Literature: Papers, references, reading notes
- Miscellaneous: Unclear or mixed content
Workflow:
- Read file from
inbox/ - Analyze content to determine category
- Extract key information
- Suggest destination
- Optionally move/process file
2. Information Extraction
Extract actionable information from unstructured content.
What to extract:
From Meeting Notes:
- Decisions made
- Action items (who, what, when)
- Key discussion points
- Follow-up questions
From Protocols:
- Procedure steps
- Required materials
- Expected outcomes
- Citations/sources
From Ideas:
- Core hypothesis
- Experimental approach
- Required resources
- Potential experiments
From Literature:
- Key findings
- Relevant methods
- Citations
- Relevance to current work
3. Destination Suggestions
Recommend where content should be integrated.
Classification Rules (from references/classification-rules.md):
| Content Type | Suggested Destination | Rationale |
|---|---|---|
| Meeting notes | notebook/knowledge/meeting_YYYY-MM-DD.md | Reusable context |
| Protocol | notebook/knowledge/protocol_[name].md | Reusable procedure |
| Experiment idea | Create new lab notebook via /research-exp | Start experiment |
| Research hypothesis | Refine with hypothesis-driven skill | Improve quality |
| Raw data file | data/raw/[filename] | Data storage |
| Literature notes | notebook/knowledge/literature_[topic].md | Reference material |
4. Integration Actions
After classification, take appropriate action.
Possible actions:
-
Move to knowledge/:
mv inbox/meeting_notes.txt notebook/knowledge/meeting_2025-01-15.md -
Create experiment:
Content suggests new experiment → Use /research-exp -
Add to STEERING.md:
Important decisions → Update priorities in STEERING.md -
Add to tasks.md:
Action items → Add to notebook/tasks.md -
Process for report:
Relevant findings → Note in ongoing report -
Archive:
mv inbox/old_file.txt inbox/archive/
Usage Workflow
Typical Usage
-
User adds file to inbox/:
User: "I put meeting notes in inbox/team_meeting.md" -
Classify content:
Assistant reads file, determines category: Meeting Notes -
Extract information:
Decisions: - Approved Exp04 (ATAC-seq analysis) - Prioritize cell cycle validation Action items: - [User] Design primers for qPCR by Friday - [Collaborator] Share ChIP-seq data next week Discussion: - Considered three-prime seq vs bulk RNA-seq - Decided bulk more appropriate for current question -
Suggest destination:
Suggested destination: notebook/knowledge/meeting_2025-01-15.md Suggested actions: - Add primer design to tasks.md - Update STEERING.md with Exp04 approval - Plan to receive ChIP-seq data -
Implement (with user approval):
Move file, update tasks.md, update STEERING.md
Batch Processing
Process multiple inbox files:
User: "Process all files in inbox/"