tylerprogramming A lightweight personal CRM. People are stored as one JSON file per person at ~/peeps/data/, and a self-contained searchable HTML page is generated at ~/peeps/index.html.
Locations
- Data (source of truth):
~/peeps/data/<id>.json— one file per person - Generated page:
~/peeps/index.html - Photos (optional):
~/peeps/assets/photos/<id>.jpg - Page generator:
~/.claude/skills/peeps/build.py
<id> is a kebab-case slug of the person's name (e.g. "Jane Smith" → jane-smith). If two people share a name, append a differentiator (john-smith-work).
Detect intent from how the user invokes the skill
| User intent | Action |
|---|---|
| Add a new person ("peeps: add Jane Smith, college roommate, birthday April 12") | Add — see below |
| Update someone ("peeps: update Mike, new phone 555-1234") | Edit — see below |
| View / open ("peeps", "show my peeps", "open peeps") | View — rebuild page and open it |
| Query ("whose birthday is coming up?", "who do I know in Austin?") | Query — read all JSON, answer in chat |
| Remove ("peeps: remove Mike") | Delete — confirm, delete file, rebuild |
Adding a person
Keep it fast — infer everything you can from the one sentence the user gave. Only ask if something important is genuinely ambiguous (e.g. you can't tell the person's name). Don't interrogate.
- Build the JSON record using the schema below. Fill what you can infer; leave unknown fields as
nullor empty arrays/objects. Always setcreatedandupdatedto today's date (YYYY-MM-DD). - Write it to
~/peeps/data/<id>.json. - Rebuild the page (see "Rebuild the page").
- Confirm to the user and show the saved record.
Schema
{
"id": "jane-smith",
"name": "Jane Smith",
"nickname": "Janie",
"photo": "assets/photos/jane-smith.jpg",
"how_we_met": "College roommate, 2008",
"location": "Austin, TX",
"tags": ["friend", "college"],
"dates": {
"birthday": "1988-04-12",
"anniversary": null,
"met_on": "2008-09-01"
},
"contact": {
"phone": "555-0142",
"email": "jane@example.com",
"socials": { "instagram": "@janie", "linkedin": "in/janesmith" }
},
"relationships": {
"partner": "Tom",
"kids": ["Ava", "Leo"],
"family_notes": "Has a dog named Biscuit"
},
"interests": ["pottery", "trail running"],
"gift_ideas": ["pottery class voucher"],
"notes": "Free-form notes.",
"created": "2026-06-13",
"updated": "2026-06-13"
}
Birthday format: ISO YYYY-MM-DD. If the year is unknown, use --MM-DD (e.g. --04-12) — the page still shows the day and "days until" but skips age. photo is optional; omit or leave null if there's no image (the page shows a colored initials avatar instead).
Editing a person
- Find the file via the slug, or Glob
~/peeps/data/*.jsonand match by name if unsure. - Read it, apply the changes, set
updatedto today. - Write it back and rebuild the page.
Rebuild the page
Run the generator after any add/edit/delete:
python3 ~/.claude/skills/peeps/build.py
It reads every ~/peeps/data/*.json and writes a fresh ~/peeps/index.html with all people inlined (works offline, no server needed). To open it after a "view" request:
open ~/peeps/index.html
Querying
For questions like "whose birthday is coming up" or "who do I know in Austin", Glob and Read the JSON files, compute the answer, and respond in chat. You don't need to rebuild the page for a pure query.
Notes
- Always confirm before deleting a person's file.
- Never overwrite an existing person on "add" — if the slug already exists, treat it as an edit or ask whether to create a differentiated id.
- The page is regenerated from scratch each build, so the JSON files are the single source of truth — never hand-edit
index.html.