pd-vault
Skill3 starsChanged 13 days agoApache-2.0View on GitHub
Use when writing memory, cards, journals, reports, notes, exports, and on «what's the number of …»: personal data become labels like ⟦тел·1⟧. RU: телефоны, почта, карты, паспорта, ПИНФЛ; «какой номер у…».
In this repository34 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- cardsSkill
- communicationSkill
- copyblockSkill
- delaSkill
- escalate-to-ownerSkill
- import-reviewSkill
- inboxSkill
- inbox-triageSkill
- iniciativaSkill
- memory-auditSkill
- mila-exportSkill
- MILAGPT-chatSkill
- mila-in-chatSkill
- mila-sessionSkill
- mila-speechSkill
- mila-tasksSkill
- mila-updateSkill
- obratnaya-svyazSkill
- oglyadkaSkill
- onboardingSkill
- owner-dmSkill
- packsSkill
- plain-wordsSkill
- raspisanieSkill
- safe-linkSkill
- self-checkSkill
- subscription-limitsSkill
- subscriptionSkill
- tasks-disciplineSkill
- tasksSkill
- topicsSkill
- utochnenie-zadachSkill
- verifySkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
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
- Very specific: line 12
After every write to memory, journals, chat cards, clients, NOW/USER/CLIENTS/DECISIONS — the pd-mask hook replaces personal data in the …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
When to use (full): Personal data vault (core layer 2a, on by default): phone numbers, e-mails, card numbers, passports, PINFL in what Mila stores or passes on become short labels like ⟦тел·1⟧; values sit encrypted in the Mila's home. Use when writing memory, cards, journals, reports, notes to other Milas, exports; when the owner asks «what's the number of …» (reveal — only to him, in his DM). RU: Хранилище личных данных (слой 2а ядра, включено по умолчанию): телефоны, почта, карты, паспорта, ПИНФЛ в том, что Мила хранит и передаёт, заменяются ярлыками; значения — зашифрованно в доме Милы. Раскрыть — только хозяину в личку.
Personal data vault
Owner's word 17.10.2026: «store personal data privately and encrypted; outside — only a short label». On by default.
What it does by itself
- After every write to memory, journals, chat cards, clients, NOW/USER/CLIENTS/DECISIONS — the
pd-maskhook replaces personal data in the file with labels (⟦тел·1⟧,⟦почта·2⟧,⟦карта·1⟧,⟦паспорт·1⟧,⟦ПИНФЛ·1⟧). Same value — same label everywhere. - Values go to
~/work/state/pd/vault.enc(AES-256 + HMAC; key~/.config/mila/pd.key, 600). Without the key the file is unreadable; a tampered file is refused.
What I do myself (the hook can't recognise it)
- Addresses, health and diagnoses, family matters, debts — I don't write them into memory as they are: «адрес — у клиента в карточке CRM», «по здоровью — к врачу». If they must be kept — only in the client's card, never in reports, examples, notes to other Milas.
- In notes to other Milas, reports and exams — labels only. Before sending text outside I run
pd_vault.py maskover it.
Reveal
python3 ~/.claude/skills/pd-vault/pd_vault.py reveal ⟦тел·1⟧ → the value. Only when the owner asks, only in his private chat, only the one he asked for. Never to other people, never into groups, never to another Mila.
Honest limit
The model reads an incoming message whole — otherwise Mila can't answer. The vault protects what Mila keeps and passes on, not the conversation itself. For full privacy of everything — the «Private mode» (encrypted home with the owner's key).
Check
pd_vault.py scan <files> — what kinds of personal data are still in plain text (no values printed); exit 1 if found.
Done if
scanover memory/, journals and cards finds nothing.- Reveal happened only in the owner's DM and only on his request.
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.

