hm-night

hm-night is a command for Claude Code from JavanC/Homunculus. It costs 0 tokens per session (2,789 once invoked), scanned A, original, MIT.

A command that runs one Homunculus evolution cycle according to the project's configuration. The cycle may include health checks, research, experiments, routing observations, evaluations, and a report.

In plain words
What is it for?
Use it to run a daily or weekly evolution cycle, check the evolution configuration, evaluate changed skills, perform configured research or experiments, and produce a report.
Why use it?
It turns the system's recurring maintenance and improvement work into a repeatable process whose depth depends on the configured tier and schedule.

Command for Claude Code

Written for Claude Code: a Claude Code command (commands/*.md). Also seen: reads .claude/ paths; mentions CLAUDE.md; positional $N argument.

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/javanc/homunculus/hm-night
Clone the repo
git clone --depth 1 https://github.com/JavanC/Homunculus

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for hm-night

README.md
[![agentmods](https://agentmods.dev/badge/commands/javanc/homunculus/hm-night.svg)](https://agentmods.dev/commands/javanc/homunculus/hm-night)
Your own site
<a href="https://agentmods.dev/commands/javanc/homunculus/hm-night"><img src="https://agentmods.dev/badge/commands/javanc/homunculus/hm-night.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,789 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00000 $0.02789
Opus 5 $0.00000 $0.01394
Sonnet 5 $0.00000 $0.00558
Haiku 4.5 $0.00000 $0.00279

Measured 6d ago against content hash a67e672f3e99, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

hm-night scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/hm-night.md · 265 lines

How it starts

The opening of the file, as written. The whole thing — 265 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/hm-night — Run One Evolution Cycle

Run the evolution pipeline. Depth depends on evolution-config.yaml (tier + schedule).

Always communicate in English regardless of user's global Claude settings.

Evolution Config

Read evolution-config.yaml from the project root at the start. If missing, default to tier: standard.

Tier controls what runs:

  • minimal — Phase 1 (health) + Phase 2 (instinct routing only) + Phase 5 (report). Skip research & experiments.
  • standard — All phases. Research: research.topics_min-research.topics_max topics. Experiments: experiments.max_per_night.
  • full — All phases + deeper research + more experiments + bonus loop.

Weekly schedule:

  • Check schedule.weekly.day (0=Sun..6=Sat, -1=every day).
  • If today matches → run weekly deep mode: full skill re-eval, goal tree mechanism review, deep health check.
  • Otherwise → run daily light mode: instinct routing + changed skill eval only.

Core Principle

Goals are stable. Implementations are diverse. When suggesting improvements, consider ALL implementation types — not just skills:

Type When to suggest
Skill Behavioral knowledge Claude should follow
Agent Task needs specialized model/tools/prompt
Hook Should trigger automatically on events
Script Automation that runs independently
Rule Claude Code behavioral constraint
Command Workflow the user triggers manually
Cron/LaunchAgent Needs to run on a schedule
MCP Needs external service integration

Always pick the right tool for the job, not default to skills.

Behavior

Run through all 5 phases systematically.

Phase 1: Health Check

  1. Read architecture.yaml
  2. For each goal with a health_check.command, run it and report pass/fail
  3. For goals without health checks, check if realized_by points to an existing file
  4. Report:
    [1/5] Health Check
          code_quality:    ✅ healthy (tests passing)
          productivity:    ⚠️ no health check defined
          ai_news:         ○ not implemented yet
    

Read the full file on GitHub · 265 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 6d ago First seen · 265 lines · 0 tokens per session scan A a67e672f3e99

Subscribe to this mod's changes

hm-night is a command published in the GitHub repository JavanC/Homunculus (15 stars, last pushed 3mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,789 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.