A collection of productive slash-command skills for Claude Code — document generation, changelog automation, project maintenance, and more.

Try before you commit. Install one skill at a time, test it, and only keep what's useful. Everything installs as individual files that are trivial to remove.

pip install python-docx python-pptx

git clone https://github.com/YOUR_USERNAME/claude-skills.git ~/claude-skills
cp ~/claude-skills/tools/convert_md_to_docx.py ~/.claude/tools/   # for /md_to_word
cp ~/claude-skills/tools/convert_md_to_pptx.py ~/.claude/tools/   # for /md_to_pptx
cp ~/claude-skills/tools/generate_letter_docx.py ~/.claude/tools/ # for /business_letter
cp ~/claude-skills/tools/generate_daily_changelog.py ~/.claude/tools/ # for /daily_summary
cp ~/claude-skills/tools/archive_old_docs.py ~/.claude/tools/     # for /daily_summary
cp ~/claude-skills/tools/add_copyright_headers.py ~/.claude/tools/ # for /add_copyright

cp -r ~/claude-skills/skills/md_to_word ~/.claude/skills/
cp -r ~/claude-skills/skills/md_to_pptx ~/.claude/skills/
# ... add only what you want

Skills are loaded at startup. Restart the CLI or IDE extension for them to appear.

See INSTALL.md for full details, project_config.json setup, and troubleshooting.

Each skill and tool is a self-contained file. To remove anything:

# Remove a skill
rm -rf ~/.claude/skills/md_to_word

# Remove a tool
rm ~/.claude/tools/convert_md_to_docx.py

# Remove everything from this repo
rm -rf ~/.claude/skills/md_to_word \
       ~/.claude/skills/md_to_pptx \
       ~/.claude/skills/business_letter \
       ~/.claude/skills/daily_summary \
       ~/.claude/skills/add_copyright \
       ~/.claude/skills/post_mortem \
       ~/.claude/skills/claude-suggestion \
       ~/.claude/tools/convert_md_to_docx.py \
       ~/.claude/tools/convert_md_to_pptx.py \
       ~/.claude/tools/generate_letter_docx.py \
       ~/.claude/tools/generate_daily_changelog.py \
       ~/.claude/tools/archive_old_docs.py \
       ~/.claude/tools/add_copyright_headers.py

Note: Project-scoped skills (.claude/skills/ inside a project) are documented by Anthropic but have several open bugs as of early 2026. Until those are resolved, install into ~/.claude/skills/ only.

What it does: Converts a Markdown file to a professionally formatted Word document.

Why you'd want it: Writing docs in Markdown but needing to deliver .docx is a constant friction point. This handles cover pages, headers/footers, inline links, watermarks, and colored border bars — no manual formatting.

Example:

/md_to_word

Claude converts MyReport.md → MyReport_v1.0_2026-04-07.docx and opens it in Word.

Requires: ~/.claude/tools/convert_md_to_docx.py, python-docx

Options:

What it does: Converts a Markdown file to a PowerPoint presentation.

Why you'd want it: Slides live in Markdown with the rest of your notes — this lets you stay in that workflow and produce a real .pptx without ever opening PowerPoint. Uses # for title slides and ## for content slides.

Example:

/md_to_pptx

Claude converts Slides.md → Slides_v1.0_2026-04-07.pptx and opens it.

Requires: ~/.claude/tools/convert_md_to_pptx.py, python-pptx

Slide format:

Options:

What it does: Converts a Markdown letter draft to a formatted business letter Word document with sender block, date, recipient, body, and signature.

Why you'd want it: Writing a formal letter shouldn't mean fighting Word's auto-formatting. Write the body in Markdown, run the skill, get a clean .docx ready to print or attach.

Example:

/business_letter

Claude generates Letter_to_ACME_v1.0_2026-04-07.docx from your draft, with your name and contact info in the sender block.

Requires: ~/.claude/tools/generate_letter_docx.py, python-docx

Options:

Files prefixed with 00_ are rendered as reference sheets without a sender block — useful for cover pages or exhibits.

What it does: Generates a comprehensive daily changelog by combining git history, open issues, and the current conversation into a narrative report, then converts it to Word.

Why you'd want it: git log alone tells maybe 50% of the story — it misses the debugging detours, the decisions made, and the issues resolved. This skill produces a full account that's useful for standups, client billing, and personal records.

Example:

/daily_summary

Claude reads Issues_Active.md, CHANGES.md, and recent git commits, writes a thorough changelog, and delivers Changelog_2026-04-07_v1.0.docx.

Requires: ~/.claude/tools/generate_daily_changelog.py, ~/.claude/tools/archive_old_docs.py, md_to_word skill, python-docx

Options (generate_daily_changelog.py):

What it does: Two things working together — automatically adds copyright headers to every new file Claude creates, and bulk-adds headers to existing files missing them.

Why you'd want it: Proper copyright headers protect your work and look professional. The automatic-on-creation behavior means you never have to think about it — every file Claude writes gets properly attributed at the moment of creation.

Automatic setup (add to your CLAUDE.md):

## File Creation (MANDATORY)
All new files must include a standard copyright header and footer:
# Copyright (c) 2025-2026 <Your Name>/<Your Organization>. All Rights Reserved.
# Project: [Name] | Filename: [filename]

Once added, Claude applies the correct comment style per file type automatically on every new file — no skill invocation needed.

Bulk backfill example:

/add_copyright

Claude scans the project, identifies files missing headers, and applies them in bulk.

Requires: ~/.claude/tools/add_copyright_headers.py, template files at ~/.claude/tools/templates/copyright/ (included in this repo)

Options:

What it does: Guides Claude through generating a structured post-mortem document after a debugging session or incident.

Why you'd want it: After a hard debugging session, the lessons learned evaporate fast. This skill captures what failed, what worked, the root cause, and the fix — while it's still fresh — in a format you can actually reference later.

Example:

/post_mortem

Claude reviews the conversation and produces a structured post-mortem covering the problem, each failed approach, the solution, lessons learned, and what to do differently next time.

Requires: No dependencies — pure prompt engineering.

Output sections:

• Header (date, summary, resolution time)
• The Problem
• What Went Wrong (numbered, with code snippets)
• The Solution (with code)
• What We Learned
• What To Do Differently (Before / During / After)
• Time Cost
• Key Takeaway

What it does: Searches existing Claude Code GitHub issues for duplicates, then auto-fills a new issue form in Chrome with your suggestion — or fills in a comment on an existing issue if a match is found.

Why you'd want it: Submitting feature requests and bug reports to the Claude Code repo is slow when done manually — find duplicates, copy titles, fill forms. This automates the tedious parts so you review and click Submit.

Example:

/claude-suggestion add a way to collapse long tool output blocks

Claude searches open issues for matches. If a strong duplicate exists, it opens that issue and drafts a comment. If not, it opens the new issue form and fills it in. You review and submit.

Requires: claude-in-chrome MCP server (browser automation), GitHub account logged in to Chrome.

Several tools auto-read a project_config.json in your project root to populate branding in generated documents:

{
  "name": "My Project",
  "prefix": "MP",
  "description": "A short description",
  "version": "v1.0",
  "logo": "assets/logo.png"
}

• Python 3.10+
• python-docx — Word document generation (pip install python-docx)
• python-pptx — PowerPoint generation (pip install python-pptx)
• claude-in-chrome MCP server — required for /claude-suggestion only

PRs welcome. Keep skills generic — no hardcoded names, paths, emails, or project-specific references. Test with at least one real project before submitting.

Released under the MIT License.