dotclaude

My personal Claude Code configuration — skills, tools, and workflows.

Structure

dotclaude/
  skills/                # Claude Code skills (SKILL.md files)
    git-workflow/        # Atomic commits with Conventional Commits
    km/                  # Unified knowledge management
    comms-archive/       # Archive Outlook/Teams into the knowledge base
    datasheet-to-md/     # PDF datasheet -> clean Markdown for LLMs
    cv-extract/          # CV PDF -> JSON Resume + portrait photo
    ris-search/          # Search Austrian legal sources (RIS)
    session-summarize/   # Summarize Claude Code session transcripts
  tools/                 # Standalone scripts and utilities
    ai-mr-review/        # AI-powered GitLab MR reviews using Claude
    tmux-claude/         # Launch Claude across repos in tmux windows

Skills

git-workflow

Enforces atomic commits after every logical unit of work using Conventional Commits format.

Knowledge Management (/km)

One skill for all knowledge management. Works in any Git repo — just type /km init to get started.

/km init                                    # Bootstrap a new knowledge repo
/km What do we know about thermal management?  # Search and summarize
/km We decided to use LTC6813               # Save content (auto-detects type)
/km update sensor-fusion: add calibration   # Modify a document
/km brain add https://gitlab.com/org/team-brain.git  # Link another brain
/km @mko PROFINET stack options             # Search a linked brain
/km @all Zephyr RTOS                        # Search all brains
/km help                                    # Show all commands

Cross-repo brains: Link other knowledge repos as Git submodules. Each brain has its own access permissions — every team member controls who can read their brain. Brains auto-update when queried (if stale >15 min) and every result includes [brain@commit] for full traceability.

Other skills

• comms-archive — archive Outlook emails and Teams messages as a daily digest in the knowledge base (requires the Microsoft 365 MCP connector).
• datasheet-to-md — convert a PDF datasheet into clean, LLM-friendly Markdown with extracted images, then review it for fidelity.
• cv-extract — extract structured data (JSON Resume) and the portrait photo from a CV PDF.
• ris-search — search Austrian legal sources via the RIS (Rechtsinformationssystem).
• session-summarize — summarize a Claude Code session transcript, with token/cost stats.

Install

# macOS / Linux / WSL
git clone https://github.com/pasrom/dotclaude.git
cd dotclaude
./install.sh

# Windows (PowerShell)
git clone https://github.com/pasrom/dotclaude.git
cd dotclaude
.\install.ps1

Both scripts link ~/.claude/skills/ entries to this repo (symlinks on macOS/Linux, junctions on Windows); use -f to reinstall. The bash script additionally registers the review alias, adds this repo to additionalDirectories, appends a # dotclaude section to your global CLAUDE.md, and supports --dir=<path> for an alternate config directory (install.ps1 does skill links only):

./install.sh --dir="$HOME/.claude-work"   # Install to alternate Claude config dir

Tools

ai-mr-review

AI-powered merge request code review for GitLab using Claude Code CLI. Reviews your GitLab merge requests locally or posts feedback directly as MR comments — including inline line-level comments on the diff.

Prerequisites:

• Claude Code CLI — installed and logged in
• glab — GitLab CLI, authenticated (glab auth login)
• Python 3

Quick Start:

# After install (see above), from any GitLab repo:
review --mr 42

Usage:

review.sh                          # Review current branch vs main
review.sh develop                  # Review against a different base branch
review.sh --list                   # List open merge requests
review.sh --mr 42                  # Review a specific GitLab MR
review.sh --mr 42 --post           # Review and post as MR comment
review.sh --mr 42 --inline         # Review with inline line comments
review.sh --mr 42 --inline --dry-run  # Preview inline comments

On first run, the script creates review_prompt.md from the included .example template. Edit it to customize the review focus for your project (these custom prompts are gitignored).

How it works:

1. Collects the diff — from your local branch (git diff) or from a GitLab MR (glab mr diff)
2. Sends it to Claude — via claude --print with a customizable review prompt
3. Displays the review — structured output with severity levels
4. Optionally posts to GitLab — as a single comment (--post) or as inline line comments (--inline)

Authentication: Uses your existing Claude Code CLI login and glab auth login. For --post and --inline, your GitLab token needs the api scope.

License

MIT