claude-uxr-skills

Personal Claude Code skills for UXR/ReOps work at Miro.

Positioning

This repo is personal R&D space for skills I use day-to-day. It complements my team's internal, team-distributed skill collections (such as the customer-insights skill and related) — it is not a fork, alternative, or competitor. Where workflows overlap, my skills hand off to the team skills via informational pointers rather than duplicating their logic. Skills here may be PR'd upstream once they've earned the keep across multiple real uses.

Skills

• research-docs — Generate UXR plan, interview guide, or insights one-pager. Bookends customer-insights (BEFORE + AFTER stages); never replicates transcript analysis or synthesis.
• prototype-audit — Read-only safety check for prototype repos against the design-team prototype playbook. Surfaces state plainly for non-technical users; never auto-fixes.
• eow-summary — Generate an end-of-week summary by scanning Claude Code transcripts, git activity across configurable scan roots, ~/dev/deliverables/ entries, and project memory. Produces a layered draft (personal log, team digest, AI Design Guild candidates) for editing before sharing.
• eod-recap — End-of-day recap of action items, deliverables, and commitments pulled from the Slack messages you sent today. Writes a dated daily log and flags deliverable-worthy items. Daily sibling of eow-summary. Requires a connected Slack MCP.

What /prototype-audit looks like

Run /prototype-audit in any prototype repo and get a plain-English report — no code knowledge required to act on it.

A repo that needs attention (real run against an internal prototype):

Prototype audit — RKO-prototype (main)

[FAIL] Branch: you're on the default branch (main). Per the playbook: never work directly on main — create a feature branch first.
[OK]   Sync: up to date with origin/main (0 behind, 0 ahead)
[OK]   Working tree: clean
[OK]   Conflicts: none
[WARN] CLAUDE.md: documented: main protection. Contradicted: merge conflicts (line 21 tells the agent to auto-resolve conflicts during rebase — the playbook says to surface conflicts and pair, never resolve blindly). Missing: force-push, self-merge.
[INFO] Last commit: 6 weeks ago by Mark Boyes-Smith

Summary: Blocking issue — you're on `main`. Tree is clean (no work to rescue), so the immediate fix is simply to spin up a feature branch before doing anything. Separately, CLAUDE.md contains a line that points the opposite way from the playbook on conflict handling — worth fixing on a future branch.

Next step: "Create a new branch called `<branch-name>` so I can start working safely."

A clean repo, safe to push:

Prototype audit — my-prototype (feature/onboarding-redesign)

[OK]   Branch: feature/onboarding-redesign (not default)
[OK]   Sync: up to date with origin/main (0 behind, 3 ahead)
[OK]   Working tree: clean
[OK]   Conflicts: none
[OK]   CLAUDE.md: documented: force-push, self-merge, main protection, conflicts.
[INFO] Last commit: 12 minutes ago by Anthony Li

Summary: All clear — safe to push and open a PR.

Next step: "Push my branch to GitHub" (then open a PR, but don't self-merge).

Severity markers: [OK] passes, [WARN] know-and-decide, [FAIL] blocking, [INFO] informational.

Try /prototype-audit in 5 minutes

Got Claude Code installed? Run these in your terminal:

git clone https://github.com/alirortb/claude-uxr-skills.git ~/dev/claude-uxr-skills
mkdir -p ~/.claude/skills
ln -s ~/dev/claude-uxr-skills/prototype-audit ~/.claude/skills/prototype-audit

Then cd into any git repo and start Claude Code:

cd ~/path/to/your-repo && claude

Once it loads, type /prototype-audit. You'll get a plain-English report like the examples above.

No prototype repo handy? cd into the cloned ~/dev/claude-uxr-skills itself — it's git-backed and works fine for a sanity check.

Install (all skills)

Once the skills installer is configured:

npx skills add alirortb/claude-uxr-skills

Or clone and run the installer — it symlinks every skill folder (any dir with a SKILL.md) into ~/.claude/skills/ and is safe to re-run:

git clone https://github.com/alirortb/claude-uxr-skills.git ~/dev/claude-uxr-skills
cd ~/dev/claude-uxr-skills && ./install.sh

Don't hand-write ln -s ... ~/.claude/skills/<skill> per skill: on macOS, re-running it once the symlink exists nests a recursive self-link inside the target dir. install.sh replaces the link in place instead, so re-running is harmless.

Design principles

Common across every skill in this repo:

• Non-intrusive. No imposed folder structures. Writes to user's current directory unless told otherwise.
• Reference, don't restate. When team standards apply, link to the canonical source rather than re-stating rules. Single source of truth, doesn't rot if upstream evolves.
• Drafts, not finals. Output is a starting point for human review, always with surfaced assumptions.
• Plain language. Audience often includes non-technical researchers and designers.

License

MIT — see LICENSE.