Update Changelog

Apply directly to docs/CHANGELOG.md:

powershell -NoProfile -ExecutionPolicy Bypass -File .ai/skills/update-changelog/scripts/update-changelog.ps1 -FromTag v0.18.9 -Version 0.19.0 -Apply

Per-commit lines only (disables automatic similarity merge):

powershell -NoProfile -ExecutionPolicy Bypass -File .ai/skills/update-changelog/scripts/update-changelog.ps1 -FromTag v0.18.9 -Version 0.19.0 -NoConsolidation

Notes:

• -Version defaults to root Directory.Build.props.
• -FromTag defaults to git describe --tags --abbrev=0.
• The script upserts ## vX.Y.Z (replaces existing section for that version or inserts after ## Unreleased).
• Default consolidation: Get-ConsolidationBucket in scripts/update-changelog.ps1 merges commits that match the same similarity bucket (for example: ShareX.ImageEditor in the subject, 2026-... blog draft series, XIP/IEIP docs, Linux install/capture documentation, IEIP/XIP proposal .md create/update under Changed, multipart / S3 multipart). Extend that function when new repetitive patterns appear.
• Always manually review for wording, missed merges, and contributor attribution (#PR, @user) before publishing.

Version Grouping Strategy

Current Unreleased Work

• Use the latest released tag as the default lower bound.
• Consolidate all commits after that tag into one heading for the target version, normally the root Directory.Build.props version.
• Do not create multiple patch or prerelease headings for the same unreleased range unless the user explicitly requests a historical reconstruction.

Historical Stable Release Reconstruction

• When rebuilding old changelog history across multiple stable releases, group entries by stable minor release boundaries.
• Use git tags and Directory.Build.props history to identify those boundaries.
• Fold patch and prerelease entries into the next stable minor heading unless a patch release was intentionally standalone.
• Retain original context if useful, for example: Feature: ... (originally v0.8.1).

Consolidation Rules

• Combine related commits that affect the same component and purpose.
• Keep different components separate unless they are part of one coherent user-facing change.
• Keep commits with external contributor attribution separate when merging would obscure credit.

Commit Entry Handling

Specific Commit Assignment

• Respect specific user requests to assign certain commits to specific versions.
• Example: "List commit 298457a under v0.11.0."
• Always verify the commit hash and subject before assignment.

Attribution

• External Contributors: Attribute Pull Requests from external contributors by including the PR number and their username.
  • Format: (#PR_NUMBER, @username)
  • Example: (#77, @Hexeption)
• Maintainer Merges: Exclude merge commits from the main maintainer (e.g., McoreD) from having explicit attribution unless they contain significant unique work not covered by other commits. The focus is on crediting other users.

Categorization

Group changes within each version using standard categories:

• Features: New functionality.
• Fixes: Bug fixes.
• Refactor: Code improvements without external behavior change.
• Build: Build system, dependencies, and packaging.
• Documentation: User, developer, proposal, and release documentation.
• Testing: Test coverage and test infrastructure.
• Performance: Performance improvements.
• Changed: Fallback for changes that do not map cleanly to the categories above.

The helper script maps infrastructure and chore-style commit types into Build unless the commit subject carries a clearer component/category signal.

Entry Consolidation to Reduce Line Count

CRITICAL: Consolidate related commits into single entries to keep the changelog concise and readable.

The automation script does this by default; agents should still edit the draft for narrative quality and any merges the heuristics miss.

Guidelines:

• Group by Component and Purpose: Combine multiple commits that affect the same component and serve the same purpose.
• Preserve All Commit Hashes: When consolidating, include all relevant commit hashes in a single line.
• Target Reduction: Aim for 30-50% line reduction by consolidating related work.

Examples:

Before (verbose):

- **Media Explorer**: Add `IUploaderExplorer` interface `(9deedf9)`
- **Media Explorer**: Implement S3 file browser `(9deedf9)`
- **Media Explorer**: Implement Imgur album browser `(9deedf9)`
- **Media Explorer**: Add navigation, breadcrumbs, search, filter `(9deedf9)`
- **Media Explorer**: Add bandwidth savings banner `(e374160)`

After (consolidated):

- **Media Explorer**: Implement provider file browsing with S3 and Imgur support, including navigation, search, filtering, and CDN thumbnail optimization `(9deedf9, e374160)`

Before (mobile features):

- **Mobile**: Add adaptive mobile theming infrastructure `(4b79ddb)`
- **Mobile**: Refactor mobile views for adaptive native styling `(a7cfb22)`
- **Mobile**: Align mobile heads with native theming defaults `(1e5f9eb)`
- **Mobile**: Complete sprint 5 mobile theming polish and docs `(30bbe98)`
- **Mobile**: Add mobile upload queue and picker `(68d97d9)`
- **Mobile**: Add mobile upload history screens `(52d6ad2)`

After (consolidated):

- **Mobile**: Add adaptive theming infrastructure with native styling polish `(4b79ddb, a7cfb22, 1e5f9eb, 30bbe98)`
- **Mobile**: Add upload queue, picker, and history screens `(68d97d9, 52d6ad2)`

Before (fixes):

- **Scrolling Capture**: Always auto-scroll to top `(1fa45f2)`
- **Scrolling Capture**: Apply workflow settings and refresh hotkeys `(971219c)`
- **Scrolling Capture**: Use current scroll position for detection `(8ac2c8b)`

After (consolidated):

- **Scrolling Capture**: Improve auto-scroll behavior and workflow settings integration `(1fa45f2, 971219c, 8ac2c8b)`

When NOT to Consolidate:

• Commits from different components (e.g., don't merge "Mobile" with "Linux Capture")
• Commits with external contributor attribution (keep separate for visibility)
• Significant standalone features that deserve their own entry

Format

Follow the Keep a Changelog format with Semantic Versioning.

## vX.Y.Z

### Features
- **Component**: Description `(short-hash)`
- **Component**: Description `(short-hash, short-hash)`

### Fixes
- Description `(short-hash)`

Workflow

Step-by-Step Process

1. Choose the Changelog Mode

   • Current unreleased work: use the latest released tag as the lower bound and the root Directory.Build.props version as the target heading.
   • Historical stable release reconstruction: use two stable release tags as the range and group output under the newer stable release heading.

2. Identify the Range

   • Default helper-script behavior: omit -FromTag to use git describe --tags --abbrev=0.
   • Explicit current-range example: -FromTag v0.21.2 -Version 0.22.0.
   • Historical stable reconstruction example: compare v0.PREV_STABLE..v0.LATEST_STABLE.

   Fallback tag listing:

   git tag -l --sort=-version:refname | Select-Object -First 10
   

3. Check Target Version Read the root Directory.Build.props <Version> property unless the user explicitly provided a historical target version.

4. Consolidate Version Headings

   • Current unreleased work: create or update one ## vX.Y.Z heading for the target version.
   • Historical reconstruction: preserve stable release boundaries, but fold patch/prerelease fragments into the stable heading unless a patch release was intentionally standalone.

5. Categorize Commits

   • Group commits into: Features, Fixes, Refactor, Build, Documentation, Testing, Performance, Changed
   • Within each category, group by component (e.g., Mobile, Linux Capture, Editor)

6. Consolidate Related Entries

   • Identify commits affecting the same component with similar purpose
   • Merge them into single, comprehensive entries
   • Preserve all commit hashes
   • Aim for 30-50% reduction in line count

7. Format and Verify

   • Ensure proper markdown formatting
   • Verify all commit hashes are present
   • Check that external contributor attributions are preserved
   • Confirm adherence to Keep a Changelog format

8. Fix Double-Encoding Mojibake After any write to docs/CHANGELOG.md, scan for corrupted UTF-8 round-trip artifacts and replace them with their correct Unicode equivalents. This occurs when UTF-8 bytes are decoded and re-encoded by the wrong tool chain. Apply this as a final step every time:

   $c = [System.IO.File]::ReadAllText('docs/CHANGELOG.md', [System.Text.Encoding]::UTF8)
   
   # Fix mojibake involving the section-sign byte pair and normalize the output.
   $c = $c -replace [char]0x00C2 + [char]0x00A7, [char]0x2014
   
   # Normalize any remaining section-sign corruption.
   $c = $c -replace [char]0x00C2 + [char]0x00A7, [char]0x00A7
   
   # Collapse 3+ blank lines
   $c = $c -replace "\r?\n", "`n"
   $c = $c -replace "`n{3,}", "`n`n"
   $c = $c -replace "`n", "`r`n"
   
   [System.IO.File]::WriteAllText('docs/CHANGELOG.md', $c, [System.Text.Encoding]::UTF8)
   

   Common victims to watch for:

   • em dash (—, U+2014)
   • section sign (§, U+00A7)
   • accented letters such as ó (U+00F3)

Example Command Sequence

# Check latest tags with version-aware sorting.
$tags = git tag -l --sort=-version:refname | Select-Object -First 10

# Check current version.
$version = Select-String -Path "Directory.Build.props" -Pattern '<Version>(.*)</Version>' | ForEach-Object { $_.Matches.Groups[1].Value }

# Preview default current-unreleased changelog section.
powershell -NoProfile -ExecutionPolicy Bypass -File .ai/skills/update-changelog/scripts/update-changelog.ps1

# Apply an explicit current-unreleased range.
powershell -NoProfile -ExecutionPolicy Bypass -File .ai/skills/update-changelog/scripts/update-changelog.ps1 -FromTag v0.21.2 -Version $version -Apply

Encoding-Safe Multi-Line Block Replacement

Why: exact-match replacement tools can fail when CHANGELOG.md contains multi-byte UTF-8 sequences that were double-encoded during a tool round trip. Use PowerShell [System.IO.File] plus Regex instead; it reads raw bytes and avoids exact-literal matching against corrupted text.

Pattern (replace all prerelease sections between two stable headings with new consolidated content):

$cl = 'docs/CHANGELOG.md'
$c  = [System.IO.File]::ReadAllText($cl, [System.Text.Encoding]::UTF8)

$newSection = @'
## v0.X.Y

### Features
- ...

### Fixes
- ...

'@

# (?s) = dotall (. matches newlines); match from first prerelease heading up to (but not including) the previous stable heading
$c = [System.Text.RegularExpressions.Regex]::Replace(
    $c,
    '(?s)## v0\.FIRST_PRERELEASE.*?(?=## v0\.PREV_STABLE)',
    $newSection
)

# Normalize the double-encoded section-sign artifact (C2 A7 → A7).
$c = $c -replace [char]0x00C2 + [char]0x00A7, [char]0x00A7

# Collapse 3+ consecutive blank lines down to 2
$c = $c -replace "\r\n", "`n"
$c = $c -replace "`n{3,}", "`n`n"
$c = $c -replace "`n", "`r`n"   # restore CRLF if the repo uses it

[System.IO.File]::WriteAllText($cl, $c, [System.Text.Encoding]::UTF8)

Key points:

• (?s) makes . match newlines so the pattern spans the whole block.
• The lookahead (?=## v0\.PREV_STABLE) stops the match at the previous stable heading; it is not consumed.
• The mojibake normalization pass ([char]0x00C2 + [char]0x00A7 to [char]0x00A7) should always run after a regex write to guard against double-encoding.
• Blank-line normalization (\n{3,} to \n\n) prevents the file from accumulating excess whitespace after sections are removed.