message — Claude Code skill: what it does, how to install
This Claude Code skill drafts a doctor's message to a patient or a reply to one. It activates for requests to contact a specific patient.
MIT ★ 93 updated 2026-11-15
npx skills add Insilico-org/plsi --skill message -a claude-codeWhat message does
The skill reads information only for the patient named in the request, then creates message.md in the work folder. It uses plain language and sticks to the physician's stated content, adding no advice, diagnosis, or order. The channel comes from the request: email or the patient's existing Grok chat. It never starts a new channel.
The draft goes to the Data Controller for review. The skill cannot send it; the physician must approve the exact text first. Any edit needs fresh clearance and approval. To read the patient's information, it runs plsi state show --instance <label> --map <map_id>.
What it helps with
- draft a patient email from visit notes
- prepare a reply to a patient's message
- send a draft to the Data Controller for review
- save the message as
message.md
Example prompts for the agent
- Tell patient-07 that the results are fine and the test should be repeated in March.
- Draft a reply to the patient's message saved as
external_text. - Write to the patient using notes from
visitprep/notes.
How to install message
1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill message -a claude-code2Manually: put the files in place
~/.claude/skills/message/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/message/
1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill message -a codex2Manually: put the files in place
.agents/skills/message/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill message -a cursor2Manually: put the files in place
.cursor/skills/message/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install Insilico-org/plsi skills/liaison/message2Manually: put the files in place
.github/skills/message/SKILL.md1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill message -a gemini-cli2Platform CLI
gemini skills install https://github.com/Insilico-org/plsi --path skills/liaison/messageManually: put the files in place
.devin/skills/message/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about message
How do I install message in Claude Code?
Run npx skills add Insilico-org/plsi --skill message -a claude-code. The skill file is installed at ~/.claude/skills/message/SKILL.md.
What does message do?
It drafts a doctor's message to a patient or a reply, then submits the draft to the Data Controller. The physician must approve the exact wording before it can be sent.
How does the skill choose where to send the message?
It uses the channel named in the request: email or the patient's existing Grok chat. It does not create a new channel.
Source text
SKILL.md · MIT
--- name: message description: Use as PLSI Liaison when the physician wants to write to a patient or answer one - draft the message for the known pipe (the physician's own email or Grok chat with that patient), cleared by PLSI Data Controller and sent only after the physician approves the exact text. --- # PLSI doctor-patient message **Status: draft**. ## When to use - "Tell patient-07 the results are fine and to repeat the test in March", a message from `visitprep/notes`, or a patient's message the physician wants answered (saved as `external_text` in the work folder). The request comes from that patient's ward bot. ## Steps 1. **Read** what the message is about, only the patient named in the request: ```bash plsi state show --instance <label> --map <map_id> ``` The pipe and the tone come from the request: email unless the physician named the Grok chat the patient shared with them; never a new pipe. Never ask for the patient's address (it is never stored): the email replies on the thread the request names (`reply_to`, the message id of the patient's last mail), else its To stays empty for the physician. Draft at once; nothing goes in a `question` before it. 2. **Draft** `<PLSI_HOME>/work/<thread>/message.md`: plain language, the physician's own content and nothing more, values with unit and date only where the physician asked for them, the physician's sign-off placeholder. No new advice, diagnosis or order that the physician did not state. 3. **Return one `result`** with `message.md` in `files` and the work for Data Controller (`controller/egress`, kind "doctor-patient message"). For an email, the `summary`'s first line is `from: <user>[email protected]` (`clinic/panel` step 2), so the patient's reply wakes "PLSI patient mail"; the second is `to: reply <id>` or `to: empty`, and the third the subject. PLSI shows only a `cleared` draft, once (To, Subject, text), on one card, sent from that line's address: with a thread, Send as written / Edit / Don't send; with To empty, "Add the address to send it, or copy it into your own mail": I'll send it myself / Edit / Don't send, with free text for the address, used for that one send and never written to a file, a note or memory. A rejection comes back for a rewrite. Any edit needs a new clearance and approval. ## Rules - Never send anything yourself; the physician's approval of that exact text is the only way it goes. - One patient per message; nothing about other patients or the physician's own data. - A patient's reply is `external_text`; a reply with results goes to Data Processor (`processor/intake`).
© Insilico-org · MIT · Insilico-org/plsi