Process File Skill

Generic file processing Skill supporting multiple file formats for parsing and intelligent archiving, fully integrated with the AkashicRecords governance system.

When to use this Skill

• User says "read", "process"

• User says "archive", "import"

• User provides file path for processing

• User wants to integrate external files into knowledge base

• User provides email, PDF, Office documents, images, etc.

Workflow

1. Initialization - Read Preferences

Check claude.md:

• Read current project's claude.md

• Look for file-handling-preferences related record

• If path found, read preferences file

If no preferences file exists:

• Ask user: "This is the first time using process-file skill in this project. Where would you like to create the file handling preferences?"

• Suggest default location: file-handling-preferences.md in project root

• After user confirmation, create file and record location in claude.md

Preferences file structure:

• Processing pattern records (by file type and content category)

• Auto processing settings (whether to allow saving without confirmation)

• Historical processing records

2. File Type Detection

Detect file type: Determine processing method based on file extension:

Type Extension Processing Tool

Email .eml mu view <filepath>

PDF .pdf markitdown <filepath>

Word .docx markitdown <filepath>

PowerPoint .pptx markitdown <filepath>

Excel .xlsx markitdown <filepath>

Image .jpg, .png, .gif, .webp, .bmp Read tool (language model direct read)

Audio .mp3, .wav, .m4a, .aac, .ogg Ask user

Video .mp4, .mov, .avi, .webm Ask user

Tool availability check:

• Check if required tools are installed before execution

• If mu not installed: Prompt Please install maildir-utils: sudo apt install maildir-utils

• If markitdown not installed: Prompt Please install markitdown: pip install markitdown

3. Content Extraction

Email (.eml):

mu view <filepath>

Extract: sender, recipient, subject, date, body

PDF/Office documents:

markitdown <filepath>

Convert to markdown format

Images: Use Read tool to directly read image, let language model analyze content:

• Identify image subject

• Extract text (if any)

• Describe image content

Audio/Video:

• Ask user for suggested processing method

• Possible options:

• Record file metadata only

• Use external tool for transcription

• Record manual summary

• Record user's chosen processing method in project claude.md

4. Content Analysis

Analyze content:

• Identify topics and keywords

• Determine content type (technical, personal, work, academic, etc.)

• Extract important information (dates, people, places, events)

Infer user intent:

• Archive for storage (long-term preservation)

• Project update (related to existing project)

• Record memo (personal notes)

• Data organization (batch processing)

Match against preferences:

• Check if preferences file has matching patterns

• If historical records exist, prioritize suggesting same processing method

5. Directory Discovery

Use akashicrecords mechanism:

• Based on content analysis results, build search query

• Scan knowledge base directory structure

• Read each directory's RULE.md to understand purpose

• Evaluate content-to-directory purpose match

Suggestion logic:

• Technical document + directory purpose "research" → high match

• Email + directory purpose "communications" → high match

• Personal photo + directory purpose "personal life" → high match

• No clear match → suggest Miscellaneous or ask user

6. User Confirmation

Present analysis results:

File Analysis Results

File: [filename] Type: [file type] Content Summary: [brief summary]

Inferred Intent: [archive/update/record]

Suggested Location: [target directory path] Reason: [why this location was chosen]

Planned Operation:

• Call add-content skill
• Format: [according to RULE.md]
• Filename: [suggested filename]

Do you approve this operation?

Wait for confirmation:

• Default requires user approval

• If auto_save: true in preferences, can skip confirmation

• User can modify suggested location or cancel

7. Execute

Call corresponding akashicrecords skill:

• Add new content → add-content skill

• Update existing → update-content skill

Format according to target RULE.md:

• Read target directory's RULE.md

• Follow naming conventions

• Apply frontmatter format (if required)

8. Update Preferences

Record this processing experience:

[Date] [File Type]

• Content characteristics: [key features]
• Target location: [actual storage location]
• Processing method: [skill used]

Learning pattern:

• Accumulate user preferences

• Prioritize suggesting same method for similar content next time

Multi-File Processing

When user provides multiple files:

Parallel Analysis

• Launch a subagent for each file

• Each subagent independently executes Phase 2-5

• Wait for all subagents to complete

Consolidated Presentation

Multi-File Processing Analysis Results

Please choose:

• Approve all
• Confirm individually
• Cancel

Batch Execution

• After user approves all, execute sequentially

• When user confirms individually, confirm each file separately

Error Handling

Tool Not Installed

Warning: Cannot process .eml file: mu tool not installed Please run: sudo apt install maildir-utils

Unsupported File Format

Warning: Unsupported file format: .xyz How would you like to proceed?

1. Try reading as plain text
2. Record file metadata only
3. Skip this file

Parse Failure

Warning: Unable to parse file content Error: [error message] How would you like to proceed?

1. Retry
2. Enter summary manually
3. Skip this file

Integration with Governance

Before operation:

• Read preferences file

• Confirm akashicrecords governance structure exists

During operation:

• Use akashicrecords skills for actual operations

• Follow target directory's RULE.md

After operation:

• Update preferences file

• akashicrecords skills automatically handle README.md updates

Examples

Example 1: Process PDF Paper

User: "Read ~/Downloads/transformer-paper.pdf"

Workflow:

• Check preferences → Find historical record "technical paper → Research/Papers/"

• Detect .pdf → Use markitdown

• Execute markitdown ~/Downloads/transformer-paper.pdf

• Analyze content → AI/machine learning topic

• Match preferences → Matches "technical paper" pattern

• Suggest Research/Papers/AI/

• User confirms

• Call add-content skill

• Update preferences file

Example 2: Batch Process Emails

User: "Archive these emails: email1.eml email2.eml email3.eml"

Workflow:

• Detect multiple files → Launch 3 subagents

• Each subagent processes in parallel:

• Parse using mu view

• Analyze sender, subject, content

• Suggest target location

• Consolidate results into list

• User selects "Approve all"

• Execute add-content sequentially

• Update preferences

Example 3: Process Image

User: "Process this photo ~/Photos/vacation.jpg"

Workflow:

• Detect .jpg → Use Read tool

• Language model analyzes image content → "Beach vacation photo"

• Check preferences → Find "travel photos → Personal/Travel/"

• Suggest Personal/Travel/2025/

• User confirms

• Call add-content (convert to descriptive markdown)

• Update preferences

Best Practices

• Always check preferences first - Prioritize historical processing patterns

• Confirm before saving - Default requires user approval

• Update preferences after success - Accumulate learning user preferences

• Use parallel processing - Leverage subagents for multiple files

• Handle errors gracefully - Provide alternatives

• Integrate with akashicrecords - Use existing skills for operations

Notes

• Preferences file path is recorded in project claude.md

• Each project can have different preferences

• Audio/video processing methods are recorded in claude.md

• This Skill does not modify files directly, operates through akashicrecords skills