Persona Check

Arguments: $ARGUMENTS

A fast, on-demand checkup of the persona layer — and a one-command way to pull every persona doc up in Sublime to edit. Lighter than /routing-health (which exercises the full model × surface × persona matrix): this needs only settings + one HTTP call, no app load, so it's quick.

Pass --no-open to run the checks without launching Sublime (e.g. headless / over SSH).

What it checks

• File health — every bot's persona_file exists, is readable, and resolves to a non-empty preamble (primary legitimately has none today).
• /chat ⇄ Telegram equivalence — /api/personas matches the bot registry; resolve_persona(id) equals the matching Telegram bot.persona; advertised capabilities match each bot's orchestrates flag. (So messaging a persona in /chat behaves identically to messaging its bot.)
• Privacy guard — none of the configured personal values (LIFEOS_PARTNER_NAME, LIFEOS_THERAPIST_PATTERNS, LIFEOS_PERSONAL_RELATIONSHIP_PATTERNS, user_name, and people_dictionary.json names) — nor absolute home paths — appear in the committed persona files. Catches a real name or personal path leaking into the open-source repo.
• Frontmatter validation (forward-compatible) — if a persona file carries YAML frontmatter, id must match the filename, voice must be a list, model a string. Silent when there's no frontmatter yet.

Instructions

Step 0: Pre-flight

!curl -s -m4 http://localhost:8000/health >/dev/null 2>&1 && echo "lifeos-api: up" || echo "lifeos-api: DOWN (the equivalence check will WARN, the rest still runs)"; [ -x "$HOME/.venvs/lifeos/bin/python" ] && echo "venv: ok" || echo "venv: MISSING"

Step 1: Run the check

From the repo root, with the project venv:

~/.venvs/lifeos/bin/python .claude/skills/persona-check/scripts/persona_check.py

It prints a [PASS]/[FAIL]/[WARN]/[INFO] line per check, a PERSONA-CHECK-VERDICT: line, and a PERSONA-CHECK-JSON: summary. Read-only — it touches no files and loads no model.

Step 2: Open the persona docs for editing

Unless $ARGUMENTS contains --no-open, open every persona doc in Sublime so they're ready to edit:

subl config/personas/*.md

(If subl isn't on PATH or you're headless, note it and skip — the checks still stand.)

Step 3: Report

Interpret the script output (don't re-run the checks yourself):

• Overall — 🟢 HEALTHY / 🟡 DEGRADED (any WARN) / 🔴 FAILING (any FAIL), from the verdict line.
• Findings — list any FAIL/WARN with its detail and where to look (table below). Mention that the docs are open in Sublime (or why they aren't).
• If everything passes, say so in a line and note how many personas/files were checked.

Failure → where to look

Failing check	Look at
file/<persona>	config/telegram_bots.json (persona_file path) and config/personas/<persona>.md
equivalence/*	config/settings.py (resolve_persona, list_http_personas, telegram_bots) — a registry/loader drift
privacy/no-personal-values	the named committed persona file — move the real name/path out into config (.env / relationship_overrides.json) and refer to it by role
schema/frontmatter	the named persona file's YAML frontmatter (id must equal the filename; voice a list; model a string)

Notes

• Read-only and cheap. Settings + one HTTP call; no app/TestClient, no model load, no probe conversations. Safe to run anytime.
• Requires lifeos-api running only for the equivalence check (degrades to WARN otherwise).
• For the deeper guarantee — that every model routes correctly across text and voice for every persona — use /routing-health (or /routing-health --live).