add
Skill2 starsChanged 4 months agoMITView on GitHub
Capture-and-file skill for a Life Agent workspace. Use this whenever the user hands over raw input — a voice-memo transcript, messy notes, a chat/phone-call log, meeting notes, a stray thought — and asks to "file this", "add this", "log this", "put this where it belongs", "capture this", "record this", or "/add". It analyses the input and merges each piece into the correct workspace file, never overwriting.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- vercel/turborepotheir own version, first seen Nov 15, 2026
- zecod/claude-queuetheir own version, first seen Nov 15, 2026
- WALKERKILLER/Look-Tongji-Notestheir own version, first seen Nov 8, 2026
- fivetaku/tikeytakatheir own version, first seen Oct 12, 2026
- tim-mccrimmon/scs-teamtheir own version, first seen Oct 11, 2026
- pilotspace/ADDtheir own version, first seen Oct 7, 2026
Used with
Projects with it also use note.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
MicrosoftSkillnote
Use this skill to record progress, partial findings, hypotheses, and page-anchored evidence while investigating a long document. Call Note aggressively — any time you read a page and confirm, narrow, or update your understanding of the user's question. Triggers include "jot that down", "save this finding", "note that page 7 says X", or (more often) implicit internal signals — you just learned something and want it recoverable later. Notes persist in the session and Review can select among them. Never paraphrase notes back from memory; write them as you go.
Also in 3 projects that use this
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- vercel/turborepo★ 31ksame name
- WALKERKILLER/Look-Tongji-Notes★ 53same name
- fivetaku/tikeytaka★ 40same name
- pilotspace/ADD★ 19same name
- maxwellsdm1867/wheeler★ 12same name
- fullymiddleaged/Clawness★ 5same name
- svyatov/handrail★ 2same name
- zecod/claude-queue★ 2same name
- tim-mccrimmon/scs-team★ 1same name
- sector137/crew★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository9 other files it gives its agents.
- 00_SELF/CLAUDE.mdCLAUDE.md
- 10_WORK/CLAUDE.mdCLAUDE.md
- 20_DOMAINS/CLAUDE.mdCLAUDE.md
- 30_KNOWLEDGE/CLAUDE.mdCLAUDE.md
- 40_SYSTEMS/CLAUDE.mdCLAUDE.md
- 50_LIFE/CLAUDE.mdCLAUDE.md
- 90_SHARED/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- sttSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 9 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 39
If it's a transcript, keep the transcriber's output as-is (don't fix errors here — the raw is the raw).
- Distinctive: line 116
Never overwrite: Always append/merge into existing content.
- Very specific: line 118
Conflicts: if new info contradicts something recorded, don't silently replace — tag it [CONFLICT] and ask.
- Distinctive: line 129
always a line on *why* it was rejected and how to apply it next time.
“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
add — capture & file
Routes any input into the right place in the workspace, following the folder map in the
root CLAUDE.md. The whole point: you brain-dump, the agent files it correctly and
tells you where everything went.
Step 0 — preserve the raw input (conditional)
Only for long or multi-topic inputs (roughly 500+ words, or 2+ distinct topics — transcripts, meeting notes, rambling memos): save the raw text verbatim before summarizing, so you can later verify nothing was dropped.
- Skip this for short single instructions ("add buy milk to my todo", "we decided on the blue logo") — go straight to Step 1. Don't create clutter.
- Location:
30_KNOWLEDGE/raw_inputs/YYYY-MM/YYMMDD_HHMM_topic.md - Format:
--- date: YYYY-MM-DD HH:MM source: {conversation / filename / transcript} input_type: {text, voice-transcript, chat-log, meeting-notes} retention: 6 months (temporary — delete when expired, no archive) --- {the raw input, verbatim — no summarizing, editing, or omission} - If it's a transcript, keep the transcriber's output as-is (don't fix errors here — the raw is the raw).
- Everything downstream (summary, routing, merge) uses this saved copy as the reference
point.
raw_inputs/is a temporary compare-buffer, not a permanent asset — files are deletable 6 months after the date in their name. Report stale ones when you notice them.
Step 1 — check the input type
- Audio file → run the
sttskill first to transcribe, then continue. - Text / notes / screenshots → analyze directly.
Transcripts — required check: proper nouns (names, companies, places) in speech-to-text are often wrong. Cross-check every name against
00_SELF/people/people.mdbefore recording it, and trust existing entries over a strange-looking transcribed spelling.
Step 2 — read the content
- Long / multi-topic (500+ words or 2+ topics) → summarize into structured sections first (topic sections + source quotes + unverified items). Don't compress on the fly — that's how details get silently dropped.
- Short / single instruction → extract directly.
Then pull out:
- People: anyone new → candidate for
people.md - Confirmed decisions ("let's go with", "confirmed", "OK") →
decisions_log.md - Rejections / taste feedback ("not this", "no", "redo") →
feedback_patterns.md - Tasks / deadlines (dated action items) →
todo.md - Self-knowledge (the owner describing their own disposition) →
character.md - Recent status / open threads →
shared_memory.md - Project or domain detail → the relevant folder
Step 3 — decide routing
| Information type | Target file |
| Recent issue, status update, handoff note | 00_SELF/memory/shared_memory.md |
Confirmed decision [OWNER] | 00_SELF/decisions/decisions_log.md |
Rejection pattern, taste filter [OWNER] | 00_SELF/decisions/feedback_patterns.md |
| Task / deadline | 00_SELF/memory/todo.md |
| New / updated person | 00_SELF/people/people.md |
Self-disposition, personality [OWNER] | 00_SELF/identity/character.md |
| Active project (has a deadline) | 10_WORK/{project}/ |
| Ongoing role / domain | 20_DOMAINS/{domain}/ |
| Research / knowledge | 30_KNOWLEDGE/ |
| Money / family / health | 50_LIFE/ |
| Genuinely unclear | shared_memory.md (and flag it) |
[OWNER] = the owner said or confirmed it. Things you inferred are [AI] — keep them separate.
Step 4 — show the plan before executing
Report the plan first:
## Update plan
| File | What I'll add | Why |
|------|---------------|-----|
| shared_memory.md | ... | ... |
| decisions_log.md | ... | ... |Apply after the user OKs. If they already said "just do it", skip the plan and execute.
Step 5 — execute (merge rules — apply to every file)
- Never overwrite. Always append/merge into existing content. Add to the matching section if one exists; create a new section if not.
- Conflicts: if new info contradicts something recorded, don't silently replace —
tag it
[CONFLICT]and ask. - Updates are additions. For an ongoing matter, add "new fact" — don't delete the old record and swap in the latest. The history is the asset.
Per-file conventions:
- shared_memory.md — newest at top; header
## YYYY-MM-DD — one-line title; bullets, terse; mark follow-ups with→ needs check/→ waiting. - decisions_log.md — confirmed only;
- [confirmed] **decision**: reason/contextunder a### YYYY-MM-DDsection. - feedback_patterns.md — category tag (
[tone] [brand] [strategy] [format] [taste])- always a line on why it was rejected and how to apply it next time.
- todo.md — convert relative dates ("Thursday", "next week") to absolute dates. Flag anything already overdue.
- people.md — new person → the file's format block; existing person → add
[YYYY-MM-DD update]to their notes. - character.md —
[OWNER]only, things the owner actually said about themselves.
Step 6 — report
Done:
- raw preserved — raw_inputs/YYYY-MM/YYMMDD_HHMM_topic.md
- shared_memory.md — [one-line] (merged)
- decisions_log.md — [one-line] (merged)
- people.md — [name] updated (merged)End with any follow-ups (waiting on X, needs confirmation).
Judgment
- Ambiguous → put it in
shared_memory.md, and mirror to a better home if one exists. - Spans multiple files → split it appropriately into each; don't fear a little duplication (important things are better in several places than lost in one).
- New person → stop and confirm before adding to
people.md. [OWNER]vs[AI]→ if unsure, record it but mark "needs check". Never fabricate.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

