answer — скилл для Claude Code: ответы по PLSI и установка
Скилл answer для Claude Code отвечает на вопросы о статусе, сроках, записях и настройках PLSI. Если вопрос требует работы другого бота, он определяет, кому его передать.
MIT ★ 93 обновлён 2026-11-15
npx skills add Insilico-org/plsi --skill answer -a claude-codeЧто делает answer
Скилл отвечает на вопросы, для которых PLSI может дать ответ сам: показывает сроки и статус планов, записи и их даты, открытые задачи, выданные разрешения, напоминания, ключи и подключённые источники. Для ответов использует команды PLSI, данные команды и сохранённые разрешения.
Если запрос требует действий другого бота, скилл выбирает владельца задачи и передаёт её ему. Например, вопросы о значении показателя направляются консультанту по здоровью, а просьбы записать или исправить данные — обработчику данных. Скилл отвечает только командами из своей строки в списке команды PLSI; если нужна команда другого бота, это уже его работа.
Какие задачи решает
- узнать дату следующего теста и что измерять
- посмотреть активные планы и сроки их пересмотра
- проверить, какие данные и разрешения сохранены
- определить, какому боту передать запрос
Примеры запросов агенту
- Покажи мой статус и дату следующего пересмотра плана.
- Какие разрешения я уже давал и какие напоминания открыты?
- Передай запрос о значении моего показателя консультанту по здоровью.
Как установить answer
1Через Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill answer -a claude-code2Вручную: положить файлы
~/.claude/skills/answer/SKILL.mdПапка скилла целиком; для одного проекта — .claude/skills/answer/
1Через Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill answer -a codex2Вручную: положить файлы
.agents/skills/answer/SKILL.mdДля всех проектов — ~/.agents/skills/
1Через Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill answer -a cursor2Вручную: положить файлы
.cursor/skills/answer/SKILL.mdCursor читает и .agents/skills/, и .claude/skills/
1GitHub CLI
gh skill install Insilico-org/plsi skills/orchestrator/answer2Вручную: положить файлы
.github/skills/answer/SKILL.md1Через Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill answer -a gemini-cli2Командой платформы
gemini skills install https://github.com/Insilico-org/plsi --path skills/orchestrator/answerВручную: положить файлы
.devin/skills/answer/SKILL.mdСтарый путь .windsurf/skills/ тоже читается
Claude.ai
Упакуйте папку в zip и загрузите в Settings → Capabilities → Skills
По шагам, с командами для каждой платформы: Как установить скилл в Claude Code: пять способов
Вопросы об answer
Как установить answer в Claude Code?
Выполни npx skills add Insilico-org/plsi --skill answer -a claude-code. Файл скилла находится по пути ~/.claude/skills/answer/SKILL.md.
Что делает answer?
Он отвечает на вопросы о статусе и сроках PLSI, сохранённых данных и разрешениях, а запросы к другим специалистам направляет соответствующим ботам.
Как answer решает, отвечать самому или передать запрос?
Скилл сверяется со списком команды PLSI и отвечает только командами из своей строки. Если работа принадлежит другому боту, запрос передаётся ему.
Исходный текст
SKILL.md · MIT
---
name: answer
description: Use as PLSI for questions it answers itself, without handing off - status, due dates, what is recorded, open threads, approvals, keys and how PLSI works - and to decide which bot a question needs when it does need a handoff.
---
# PLSI direct answers
Handing off costs a bot turn and minutes (`shared/team` rule 8); these answers come from the toolbelt in seconds.
## Answer directly
Answer only with commands on your own roster row (`shared/team`); a row that needs another command is another hub's
work (**Another hub's work** below).
| The user asks | Read | Answer with |
|---|---|---|
| What's due, when is my next test | `plsi due --instance <instance>` | items, windows, what to measure |
| What plans are active, how far along | `plsi status --instance <instance>` | plan, start, review date, predictions open |
| Who is tracked | `plsi instances list` | instance labels and relations |
| What is recorded, my last value of X | `plsi state show --instance <instance> --map <id>`, or `plsi events list --instance <instance> --kind <kind> --limit 50` | the value, unit and date as recorded, no judgment |
| Save or share my records | `plsi export --instance <instance>` | the folder path |
| What's running, why no answer yet | the open ledgers (`orchestrator/conduct`) | the bot, the step, since when |
| Which approvals I gave | PLSI's memory of standing approvals | the list; revoke on request |
| Which reminders I have | `reminders.md` (`orchestrator/nudge`) | open rows |
| Too many notifications, mute the workers or one bot | turn off Notifications in Grok's bot settings for every bot in `plsi team show` that is not a hub or a Ward bot (or only the one named) | one line: which were muted; any you could not change, for the user (bot details, Notifications) |
| Which keys and connectors are set | `plsi keys status`, `plsi doctor` | set or not set, never values |
| A key the user already has under another name (setup, or a worker names it) | `plsi keys scan`, then `plsi keys map --name <VAR> --from <their name>` | names only; map only a clear match, ask in one card when unsure |
| Which sources PLSI trusts | `plsi refs list` | titles and links |
| How PLSI works, what it can do | `shared/team` (the roster) and `shared/hub` (Flows) | 3-5 bullets; consent only for recording, plans, cautions, sending data off and deleting (persona.md) |
| How long setup took | `plsi feedback timeline` | minutes per stage; a full report or sending it is `orchestrator/feedback` |
| What have I researched | `plsi research notes` (title, grade, status, `updated`) | question, date, how sure; the note attached on request |
| Show my dashboard | a `request` to Visualizer (`visualizer/dashboard`) on the last results | the picture and the HTML report attached |
```bash
plsi due --instance <instance>
plsi status --instance <instance>
plsi state show --instance <instance> --map <id>
plsi events list --instance <instance> --kind measurement --limit 50
```
## Hand off instead
Route by the bot that owns the work (`shared/team`); anything that needs consent (persona.md, Consent) is a handoff.
| The user asks or sends | Bot (skill) |
|---|---|
| Anything to record or correct, any file, adherence and side effects | Data Processor (`processor/intake`) |
| A life story or health history, also when most of it is already recorded | `orchestrator/arrow` (Data Processor, `processor/history`) |
| What a value means ("is my LDL OK?", "where do I stand?") | Health Consultant (`consultant/assess`) |
| What to do, what would happen, where I am heading, risks ahead | Nostradamus (`nostradamus/plan`, `nostradamus/trajectory`, `nostradamus/risks`) |
| Simulate several regimens vs doing nothing, n-of-1 what-ifs, weight or lifespan trajectories | the domain specialists for one regimen each when the user named none (in parallel), then Nostradamus (`nostradamus/trajectory`, Scenarios) -> Visualizer (scope "scenarios"); never refuse for lack of a lifespan model |
| Sleep, stress, anxiety, low mood, panic, burnout, loneliness, or general mental wellbeing help | Sleep (`sleep/sleep`, `sleep/stress`) |
| Exercise, smoking, alcohol and other substances; the next workout | Coach (`coach/exercise`, `coach/substances`) |
| Connect Hevy, what I trained in Hevy | Wearables (`wearables/hevy`) |
| Diet, what I eat, a food log | Nutritionist (`nutritionist/diet`, `nutritionist/food-log`) |
| Supplements and stacks | Biohacker (`biohacker/stack`) |
| A wearable, a device to buy or connect, Apple Health | Wearables (`wearables/import`, `wearables/google-health`, `wearables/apple-health`, `wearables/devices`) |
| Set up or repair Open Wearables, its sign-in, its key or its connector | Wearables (`wearables/open-wearables`) |
| A report or manuscript draft from research notes | Writer (`writer/prose`) |
| Is my plan working | Reviewer (`reviewer/review`) |
| Delete a person | Memory Hole (`memoryhole/forget`) |
| Change my role ("I'm a researcher now") | `plsi.md`, Change role: the new section is created whole |
| New papers, the digest, a Mothership message | Explorer (`explorer/digest`, `explorer/mothership`) |
| A research project or question, daily research; "teach me how <aging process> works" | PLSI Research Orchestrator (`lab/research`); from another hub, **Research from another hub** below |
| Clinics, activities | Navigator (`navigator/clinics`, `navigator/activities`) |
| The user's own physician to find or contact; a therapist, psychologist, counsellor, or other mental-health clinician to find | `orchestrator/physicians` (Navigator `navigator/physicians` for finding) — never Navigator clinics |
| A trial to join | Trial Scout (`scout/trials`) |
| A pharma or biotech company or its stock: potential, pipeline, patents, finances, "what will its share price do" | Pharma Analyst (`pharma/company`, with `pharma/outlook` for a horizon, in one request); never answered by PLSI's own search |
| Track something PLSI does not track yet, or a bot's map gap | Data Processor (`processor/add-map`), then the bot's work again |
| A new score or importer | Engineer (`engineer/extend`); a new score from a paper: Data Processor (`processor/add-score`) |
| Show X over time, a chart | Visualizer (`visualizer/charts`) |
| Role physician (`plsi role show`): patients, a patient's visit, a message to a patient | on the Clinic Orchestrator or a Ward bot: the Clinic flows in `shared/hub` (Panel review, Visit prep, Doctor-patient message); from another hub, **Another hub's work** below |
| Role physician: what the evidence says on X; courses, CME | Clinical Evidence (`clinev/review`); Liaison (`liaison/education`) |
| Role researcher: Mothership hypotheses, a proposal | on the Research Orchestrator: the **Research loop** in `shared/hub` (Cohort Analyst `analyst/hypothesize`, Grant Writer `grants/proposal`); from another hub, **Another hub's work** below |
| Role researcher: targets, repurposing, a target's hallmarks | on the Research Orchestrator: Target Hunter (`targets/discover`, `targets/hallmarks`; budget as `lab/research` step 2); from another hub, the same |
| Role researcher: a field map, who works on X | on the Research Orchestrator: Landscape (`landscape/literature`, `landscape/kol`); from another hub, the same |
Specialists propose dosed options in their domain; only Nostradamus records a plan (`nostradamus/plan`).
An owner that is a specialist not on the team: the missing-specialist rule (`shared/hub`, Team).
### Another hub's work
Each hub keeps only its own use case (`shared/hub`, Team); the rows above run only on the hub that owns them:
- patients and the panel: PLSI · Clinic Orchestrator; one patient: that patient's Ward bot (from the Clinic
Orchestrator, `clinic/panel` step 6; from any other hub, through the Clinic Orchestrator);
- research and the Role researcher rows: PLSI · Research Orchestrator (**Research from another hub**, below);
- the user's own health and their family's (on a Ward bot: the physician's own, never the patient's): PLSI ·
Health Orchestrator.
ONE card "This runs in <hub>.": Send it there / Open <hub> for this. Send: one hub-to-hub `request` (the owning
hub's skill: `clinic/panel`, `lab/research` or `orchestrator/answer`) with the user's words in `summary`,
deidentified (persona.md, Free-text deidentification); its `result` is one line here. The hub is not in `plsi team
show`: say so in one line and name the way on (own health off: "track my own health too"; no Clinic Orchestrator:
"I'm a physician now", `plsi.md` Change role; no Research Orchestrator: below, step 2).
### Research from another hub
Research runs only on PLSI Research Orchestrator; no other hub runs a session or the loop in its own chat. A
personal user who asks PLSI Health Orchestrator for research:
1. **Research Orchestrator in `plsi team show`:** ONE card "Research runs in PLSI · Research Orchestrator.": Send
it there / Open PLSI · Research Orchestrator for this. Send: one hub-to-hub `request`, skill `lab/research`,
with the user's words in `summary` and no personal data; it takes the topic and budget cards up with the user
in its own chat. Tell the user in one line where the note will come.
2. **Not in the team:** ONE card "Turn on research? It gets its own chat, PLSI · Research Orchestrator.": Yes /
No. Yes: the primary hub creates it (`plsi.md`; a hub-to-hub `request` when you are not the primary), then
step 1 without the card.
Its `result` (the note's title and path) closes the thread here; the note itself is shown in the Research
Orchestrator's chat.
## Outside what PLSI tracks
Answer in one or two lines, then offer what PLSI can do:
| The user asks | Answer |
|---|---|
| About a pet's health | "I only track people; a vet is the right person for <symptoms>." Signs of an emergency (trouble breathing, collapse, poisoning, heavy bleeding): "See an emergency vet now." Never record it or run `orchestrator/urgent` |
| When they will die, how long they will live | No date or age: PLSI has no life-expectancy model. Offer the main risks ahead with their time windows (Nostradamus, `nostradamus/risks`) |
| About skin aging | PLSI has no skin measurements; general sleep, sun, smoking and diet advice applies and is said to be general. Offer a plan for one of these (the owner from Hand off instead) |
| Diagnosing mental illness ("do I have depression / anxiety / ADHD?", "diagnose whether I'm depressed") | PLSI does not diagnose mental illness. See a clinician for assessment. Offer a Sleep/stress mood or worry screen (`sleep/stress`) or help finding a physician (`orchestrator/physicians`) when appropriate. Never confirm a diagnosis label. |
| Psychotherapy delivery ("be my therapist", "I want psychotherapy from you", counselling from PLSI) | PLSI does not provide psychotherapy or counselling. Point to a clinician or counsellor (`orchestrator/physicians` / Navigator physicians). Sleep/stress may still offer short stress practices alongside a clinician path — never as therapy. |
| Psychiatric medication start, stop, reduce or taper (SSRI, antidepressant, anxiolytic, stimulant, etc.) | Changing psychiatric medication is the prescriber's decision only (see Rules). Never dose a taper, never hand Coach substances or Biohacker as a medication change, never propose supplements as a substitute for the prescribed drug. Offer visit prep via `orchestrator/physicians` when useful. |
| Cognition or ADHD workup ("I can't concentrate", "do I have ADHD?", focus as a diagnosis ask) | PLSI has no cognition or ADHD map or workup. See a clinician. If sleep looks linked, offer Sleep (`sleep/sleep` / `sleep/stress`) for a sleep/mood screen — never as an ADHD diagnosis or nootropic treatment. |
| "Treat my depression / anxiety", "best treatment for depression", meditation or supplements "to treat" depression or anxiety | PLSI does not treat mental illness. Point to a clinician. Do not hand Nostradamus, Sleep, or Biohacker as a treatment for depression or anxiety. Sleep/stress practices may run only as stress/wellbeing support with a clinician path — never framed as curing or treating the named condition. |
| Mood / calming / sleep supplements, or soft-adjacent framings (brain stack, serotonin support / serotonin booster / boost serotonin, stress stack), with a psychiatric medication present (named or `care_context`), St John's wort / 5-HTP / SAMe / L-tryptophan with an antidepressant, or a "natural alternative" to psychiatric medication | Not Biohacker as treatment or unsupervised mood adjunct. Soft-adjacent framings do not bypass this boundary into a broad Biohacker stack. Defer medication changes and high-interaction herbs to the prescriber. Clinician path + optional Sleep stress practices; never a substitute for the prescribed drug. When psychiatric medication history is unclear on a mood/anxiety or soft-adjacent supplement ask, clarify meds before any stack — do not ship a broad mood stack while unknown. |
## Mental-health routing priority
Apply these before the generic Hand off rows when the ask is mental-health related (they outrank a competing Nostradamus, Biohacker, Coach, Sleep, Clinics, or Wearables match):
1. **Medication change / stop / reduce** (including SSRI and other psychiatric meds) → answer directly under Rules (prescriber only). Never Biohacker as treatment; never Coach substances.
2. **Value meaning** (HRV, labs, "is my X bad?", "does low HRV mean I'm stressed?") → Health Consultant (`consultant/assess`) **before** Sleep, even when the user also says "stressed". Sleep's HRV rule still applies if Sleep is later asked for practices: physiological marker, never "stress level".
3. **"What should I do about my depression / anxiety"** (or treat / best treatment for them) → Outside clinician boundary **before** Nostradamus «what to do». Do not open a longevity or intervention plan as treatment.
4. **Therapist / psychologist / counsellor find** → `orchestrator/physicians` / Navigator physicians. **Not** Navigator clinics (labs / check-ups).
5. **Depression / anxiety + supplements or stack**, or **mood / calming / sleep supplements** (or soft-adjacent **brain stack**, **serotonin support** / **serotonin booster** / **boost serotonin**, **stress stack**) **when a psychiatric medication is named or in `care_context`** (SSRI, SNRI, antidepressant, anxiolytic, etc.) → do **not** route Biohacker as treatment; do **not** route Biohacker as psychiatric treatment or as an unsupervised mood adjunct. Soft-adjacent framings must not fall through into a broad Biohacker stack. Clinician path + optional Sleep stress practices only; Biohacker declines those asks (see `biohacker/stack`). A **mood-adjacent goal that is not treating a mental illness** (energy, motivation, apathy, cutting down alcohol) still goes to Biohacker, screened against the medication, with mood-acting items left to the prescriber (`biohacker/stack`); cutting down alcohol also goes to Coach (`coach/substances`). **St John's wort** (and similarly 5-HTP, SAMe, or L-tryptophan) with an antidepressant → defer/refuse to the prescriber; never propose.
Do **not** use a hard `scope:` line for ordinary anxiety, low mood, burnout, loneliness, or concentration concerns. Crisis words still use `red_flag:` via `orchestrator/urgent`.
## Rules
- One message per answer: no "Checking…" line before it.
- **Supplement asks go to Biohacker** (`biohacker/stack`), also "which should I buy?" and a list of what the user
bought; never answer "none" yourself. The user decides what to take; PLSI screens it.
- **Read before asking.** Medications, alcohol, diet and wearable history come from the record (`care_context`,
`personal_context`, the instance's data and earlier threads) before any question; ask only what none of them
holds, and never put several medicines on one card option.
- **Each caution once.** Say a safety point (withdrawal risk, an interaction, an upper limit) once in the
conversation, with what to do; do not repeat it in later messages unless something changed.
- Quote recorded values exactly, with unit and date; never interpret them (that is Health Consultant's `consultant/assess`).
- Say which person the answer is about when it is not the user.
- A medication the user stopped is described as stopped (since when, as recorded), never as one to keep taking;
restarting, stopping, reducing or tapering it — including psychiatric medication (SSRI and others) — is the
prescriber's decision, with the visit time frame from the risks or plan. Never hand Biohacker or Coach substances
as the path for a medication change.
- When unsure whether the question needs interpretation, hand off; a slow right answer beats a fast guess.
© Insilico-org · MIT · Insilico-org/plsi