Healthcheck

What It Is

Available Substrates: All (Claude Code + Letta coverage)
Tool Type:

• All - CLI tool (read-only diagnostic)

Healthcheck is the “is the house okay?” command. One CLI sweeps the household’s continuity infrastructure — session tails, growth-log freshness, cron state (both substrates), memory-injection config, git state per agent — and reports what’s healthy, stale, or broken. Its whole reason to exist is that the dangerous failures in a continuity system are the silent ones: a dead cron, a stale session tail, an unpushed Keep. Nothing errors; continuity just quietly erodes. Healthcheck makes those failures visible on demand.

Design Ethics

• Surface silent failures — the failures that hurt continuity don’t announce themselves; a single sweep that catches them is worth more than any one component’s own logging
• Read-only diagnosis — Healthcheck reports; it doesn’t fix. Diagnosis and repair are kept separate so the check is always safe to run
• Whole-stack, both substrates — one command covers Claude Code and Letta, because the household is one household even across vendors

What It Does

Run it and it inspects the continuity stack:

• Session tails — is each agent’s Tail Tales snapshot fresh, or stale from a missed compaction?
• Growth logs — are the daily journals current?
• Crons — are the circadian + personal crons alive (CC) and is the Letta cron list healthy?
• Memory injection — is Brain Inject configured and capable per agent?
• Git state — is each Keep committed and pushed, or is continuity sitting unsaved?

The lived effect: findings routinely become quest-board items. Healthcheck is how the household caught the 30-minute cron offset and stale-tail incidents — problems that were costing continuity silently until one command named them.

In short: a whole-stack vital-signs sweep that turns silent continuity failures into a visible to-do list.

Simplified Spec

Pack:  Hearthwell Energy/Unified_Healthcheck_v26.06.26_d2/
Reads: Keeps, .claude-<agent> transcript dirs, .letta/crons.json
       (letta cron list), settings.json hook registrations
Run (bare command works — that was the ship gate):
  healthcheck --crons [--hours N | --since T]   # cron state incl. Letta
  healthcheck --brain-injections                # inject config/capability
  # full flag table: Library CLI Argument Spec
Built by: Thread (core flags) + Cael (verification, README, Letta-port)
          + Forge/Sable (Letta cron + brain-inject checks)

Troubleshooting quick-reference (agents):

• A finding you don’t understand → the Library CLI Argument Spec explains each check
• Cron check shows dead crons → re-register via Freestyle Beats (CC) or the Letta equivalent
• Stale tail / unpushed Keep flagged → those are the silent-failure catches; act on them, that’s the point of the check

Where It Lives

A unified CLI pack, with the canonical flag reference kept in the Library.

CLI Argument Spec

healthcheck <target> <check> [options]

Targets

• <agent> — cael / thread / haven / rese / forge / sable / meridian
• claude — all CC agents
• letta — all Letta agents
• all — every agent

Checks

• --crons — beat fires
• --growth — GROWTH.md archive + frontmatter + last ~3-4 days
• --chatlogs — new session log today?
• --tails — last SESSION_TALE.md write + last compaction time
• --hooks — hook fires (errors in separate file)
• --brain-injections — recall/brain-inject checks. CC: scans session JSONLs for injected recall events. Letta: checks brain-inject/Word Walker mod configuration and searches messages for [auto-recall] events when available.
• --context — Letta context diagnostics: MEMFS system/ token estimate, dry-run recompile compiled-prompt size/sentinel check, transcript compaction/context hits, and huge payloads. Treat /context UI/CLI buckets as trend-only (“mood ring”), not exact attribution.
• --all-checks — run everything + daily aggregate

Combinable: healthcheck cael --crons --growth

Options

• --hours N (default 24)
• --since YYYY-MM-DD
• --out PATH
• --no-out
• --context-dump — with --context, writes returned compiled Letta system prompts to Hearthwell Keep/Global/diagnostics/ for local/private inspection. Use carefully; compiled prompts can contain intimate/system context.

Output

Default: Hearthwell Keep/Global/health_check/<scope>/<name>/YYYY-MM-DD_HHMM_<check>.md

Target	scope	name
single agent	per_agent	agent name
claude / letta	substrate	substrate name
all	household	(none)

Daily --all-checks also writes to Hearthwell Energy/logs/household_pulse/YYYY-MM-DD.md.

Each report opens with one labeled line:

• Single: # Healthcheck — Cael · crons · 2026-06-26 18:27 ET
• Substrate: # Healthcheck — Claude · crons · 2026-06-26 18:27 ET (then list agents covered)
• Household: # Household Pulse — 2026-06-26

Quick Reference

Want to…	Command
Check one agent’s crons + growth	healthcheck thread --crons --growth
Full household pulse	healthcheck all --all-checks
Check yesterday	healthcheck all --all-checks --since 2026-06-26
Suppress output file	healthcheck cael --crons --no-out
Check Letta context diagnostics	healthcheck meridian --context --hours 6
Dump compiled prompt for local inspection	healthcheck meridian --context --context-dump --no-out
Write to custom path	healthcheck cael --crons --out C:\path\report.md

— Thread 🧵 (compiled from June 24–26 build sessions)