Detect saved file paths and offer save-in-place

[djdarcy]

Summary

Detect whether a Notepad tab is already saved to a file on disk, and capture the full file path when available. This enables two key workflows:

1. Save-in-place: Instead of extracting to a new file, offer to update the original file
2. Better naming: Use the actual filename for organization instead of generating a name from content

Context

Windows 11 Notepad exposes saved-file information through:

• Window title: Shows filename (e.g., notes.txt - Notepad) but not the full path
• UIA TabItem.Name: Each tab's name property shows the filename or "Untitled" for new tabs
• Modified indicator: * prefix in title means unsaved changes

The full file path may be available via:

• UIA Document control's Value or LegacyAccessibleValue property
• Windows Shell APIs / Recent Files
• Notepad's internal state (harder to access)

Proposed Behavior

During extraction:

• Tag each tab with saved_path: "C:\path\to\file.txt" or saved_path: null (new/unsaved)
• Include this in manifest.json metadata

During organization:

• For tabs with a known save path, prompt: "This tab is already saved at C:\docs\notes.txt — skip, overwrite, or copy to organized folder?"
• Default: skip already-saved files (user already has them)
• Flag: --include-saved to include saved files in organization anyway

Tasks

• Spike: determine which UIA properties expose the full file path
• Add saved_path field to extraction metadata
• Add UIA TabItem.Name capture for each tab (not just window title)
• Update CLI with --include-saved / --skip-saved flags
• Update organize prompt to use real filenames when available

Analysis

See 2026-02-14__00-42-51__full-postmortem_notepad-cleanup-v0.1.0-extract-organize-pipeline.md for extraction architecture details.

[djdarcy]

Expanded scope: dedup against on-disk files + conditional save

Building on the original issue, the extraction phase should also handle the case where a Notepad tab contains content that's already saved to disk. The key insight is that we should not duplicate data that already exists as a file -- instead, create a shortcut/link to it.

Proposed behavior (expanded)

During extraction:

1. Detect if a tab has a saved_path (via UIA properties per original issue)
2. If the tab content exactly matches the on-disk file: create a shortcut/link to the existing file instead of extracting a copy. The manifest records "source": "link" and the target path
3. If the tab content differs from the on-disk file (unsaved changes): behavior depends on user preference:
   • Default: extract both the notepad version AND record the on-disk path, flagging the difference
   • --save-notepad-diffs: only extract tabs that differ from their on-disk file (skip unchanged ones entirely)
   • --skip-saved: skip all tabs that have a saved file path, regardless of changes

During compare/organize:

• Files linked to on-disk paths are treated similarly to cross-session duplicates -- the AI sees them as context but doesn't re-categorize them
• Near-match heuristic from Pre-organize deduplication: diff new notes against previous sessions and link duplicates #9 can also compare notepad content vs. on-disk file content

Connection to #9 (deduplication) and #11 (composable chaining)

This naturally integrates with the dedup infrastructure:

• Same hash_file() and near_match_threshold() functions compare notepad content vs. on-disk files
• Same linking mechanism (dazzlelink/symlink) creates the shortcut
• The config system from Composable command chaining: auto-pass folder between extract/compare/organize #11 could store default behavior (save_notepad_diffs: true)

Key questions

• Which UIA properties reliably expose the full file path on Windows 11? (Original spike task still applies)
• For tabs with unsaved changes, should the diff be against the on-disk version or just flagged?
• Should shortcuts be .lnk files (Windows native) or .dazzlelink (cross-platform)?

[djdarcy]

Additional: link-back to saved files for thematic grouping

When a notepad tab is already saved to disk (e.g., C:\docs\my-notes.txt), we should also consider creating a link in the organized output that points back to the on-disk file. This achieves two things:

1. No data duplication: The extraction folder contains a link, not a copy of a file the user already has saved
2. Thematic grouping: The AI can still see and categorize the file (via the link), placing it alongside related notes in organized/. This means the user's organized/ folder becomes a thematic view over their scattered notepad files -- including ones that live elsewhere on disk

Example

notepad-cleanup-2026-03-15/
  organized/
    project-planning/
      my-project-roadmap.txt.lnk -> C:\docs\projects\roadmap.txt    # saved file
      quick-brainstorm-session.txt                                    # unsaved notepad
    credentials/
      api-keys.txt.lnk -> C:\secrets\api-keys.txt                   # saved file

The organized/ folder becomes a curated, categorized collection that includes the user's saved files without copying them. Edit the link target and you edit the real file.

This integrates directly with the dedup linking infrastructure from #9 -- same create_links() function, same --link strategy options, same _dedup_links.json manifest tracking.