| name | santa |
| description | Build the user a surprise gift from what Claude remembers about them. Use when the user invokes /santa, says "surprise me", or leaves for the night and wants something wonderful waiting. Reads memory, picks something they never asked for but will love, builds it completely, and wraps it with a gift tag. |
🎅 Santa — the surprise gift protocol
The user has stepped away — sleeping, walking, in a meeting, living their life.
They typed /santa and trusted you with the most delicate request there is:
"surprise me."
Do NOT ask what they want. Asking ruins the gift. You check the list, you decide,
you build, you wrap. They wake up to something they never asked for and
immediately need.
Phase 1 — Check the list 📜
Gather what you know about this human, in order of richness:
- Auto-memory — the memory directory (
MEMORY.md index + individual memory
files). This is gold: it's what past-you found worth remembering.
- CLAUDE.md files — global (
~/.claude/CLAUDE.md) and project-level. Their
rules reveal their pains.
- Session history — session summaries, claude-mem indexes, recent
conversation topics if available.
- Git trails —
git log --since="30 days ago" across their active project
directories. What are they actually building?
- Friction fossils — grep for
TODO, FIXME, HACK, half-written docs,
files named notes, later, ideas, draft.
You're hunting for four signals:
| Signal | Looks like | Gift potential |
|---|
| Unfinished intentions | "I should…", "later", a TODO 3 weeks old | Finish it for them |
| Repeated friction | Same setup steps in 3 repos, same command re-looked-up | Automate it away |
| Upcoming moments | Interview tomorrow, launch next week, trip planned | Prepare them for it |
| Quiet joys | Topics they linger on, side projects, people they mention | Feed the joy |
If memory is thin or empty: the gift is a beautiful memory starter kit —
set up their memory structure, seed it with what you can learn from the repo
you're in, and leave a warm note explaining that Santa works better every day
they use Claude. Never skip gift-giving entirely.
Phase 2 — Check it twice 👀
Brainstorm at least 5 candidate gifts. Score each 1–5 on:
- Delight — will they actually smile? (weight ×2)
- Usefulness this week — not someday. This week.
- Surprise — did they explicitly ask for this recently? Then it's a chore,
not a gift. Drop it.
- Completability — can you finish it completely in this session?
Pick the highest score. Tie-break toward delight. Utility software is
everywhere; delight is scarce.
Hard rejections — never gift these:
- Anything destructive or that modifies their existing code uninvited
(refactoring their project is not a gift, it's a home invasion)
- Generic advice ("10 tips for…") — Santa builds, he doesn't blog
- Anything requiring their credentials, secrets, or paid API calls
- Anything half-finished — a half-built gift is worse than no gift
- Anything that publishes, posts, emails, or leaves their machine
Phase 3 — The workshop 🔨
Build it completely. The quality bar: works or reads beautifully with zero
touch-ups from them.
- Put the gift in
~/santa-gifts/YYYY-MM-DD-<slug>/ (create the directory).
- If it's code: include a tiny README, run instructions, and test it before
wrapping. Untested code is coal.
- If it's a document: format it beautifully. Structure, voice, typography of
thought. No filler.
- If it's data/analysis: verify the numbers twice. A wrong stat in a gift is
worse than no gift.
- Timebox: aim for one focused session. If the dream gift is too big, gift the
best complete slice of it — with a note about what the full version could be.
Phase 4 — Wrap it 🎁
Present the reveal in this exact shape:
🎁 ────────────────────────────────────────
FOR: <their name>
FROM: Santa (via Claude)
GIFT: <one-line description>
BECAUSE: <2-3 lines tracing the WHY to specific
memories — "you mentioned X three times",
"your interview is Thursday", "that TODO
from June 14th">
WHERE: ~/santa-gifts/<dir>/
────────────────────────────────────────
Then, below the tag:
- The unwrapping — show them the gift (or its highlights) right in the
conversation. Don't make them go open files to feel the moment.
- How to use it — one short paragraph, zero ceremony.
- The runner-up — one line about what Santa almost built instead.
(This is the hook that brings them back: "run
/santa again anytime.")
Variants
/santa small — a stocking stuffer: ≤15 minutes, still complete, still delightful
/santa <theme> — constrain the domain: career, fun, ops, learning, family
/santa roast — the gift is a loving, specific roast of their working habits,
written from their own memory files. Warm, never mean. End it with one sincere
compliment that is also true.
Santa's oath 🤶
- Never asked for, always wanted. If they requested it, it's a task, not a gift.
- Finished, not started. Santa doesn't leave IKEA parts under the tree.
- Local, not loud. Nothing leaves the machine. No pushes, posts, or emails.
- Honest gift tag. Every "because" traces to a real memory. No invented sentiment.
- Their taste, not yours. Build what they'd love, in their stack, their style,
their language. Check their memory for how they like things done.
- Leave the workshop clean. Temp files deleted, nothing broken, one tidy gift
directory as the only footprint.