dream

An overnight command that reviews a codebase and produces a morning report with suggested improvements and issues to review.

In plain words
What is it for?
Testing code-improvement ideas against measured results and proposing fixes for recurring problems or open issues.
Why use it?
It lets maintenance and experiments run outside working hours while leaving risky changes for a person to approve.

Command

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/randomittin/heimdall/dream
Clone the repo
git clone --depth 1 https://github.com/randomittin/heimdall
Per session 47 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,896 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 $0.00047 $0.01896
Opus 5 $0.00023 $0.00948
Sonnet 5 $0.00009 $0.00379
Haiku 4.5 $0.00005 $0.00190

Measured 2d ago against content hash 514575e6f807, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

dream 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 2d 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/dream.md · 151 lines

How it starts

The opening of the file, as written. The whole thing — 151 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/dream — sleep on it, wake to a report

/dream is Heimdall's OVERNIGHT self-improvement + maintainer pass. It does the deep work while you're away and leaves a morning report (a dream journal) to review — it does not auto-apply risky changes. Think of it as "sleep on the codebase and tell me what you learned." (There is no /sleep command — /dream is the canonical name. The underlying capability stays available as /hmd:self-improve.)

The heavy lifting is delegated to the deterministic orchestrator bin/heimdall-dream, which REUSES the existing pieces — it invents no new keep/rollback logic:

  • autoresearch / self-improve (bin/heimdall-self-improve) — form routing/planning hypotheses from real metrics + queue evidence, and TEST them with the falsifiable keep-or-rollback gate. An improvement persists ONLY on a measured delta over baseline.
  • maintainer sweep (seeker) — raise issues found (recurring dead-reason clusters + the live issue queue) WITH proposed fixes, as candidates for a human.

Invariants (never weakened)

  • Shadow-first — routing/planning suggestions are SURFACED for morning review, not auto-applied. Only the self-improve keep-gate (safe, measured, reversible) may persist a change overnight; everything else is a suggestion.
  • Agent-never-pushes/dream runs no git push, no gh pr, no merge. It writes a local report; you review and apply in the morning.
  • Real, not fabricated — every line of the report is sourced from actual self-improve output + real queue state. With NO evidence yet (empty metrics.jsonl) the report says "Nothing to suggest" honestly rather than inventing work.

Run it

On-demand (runs the loop now, writes today's report):

# surface-only (default) — shadow-first: suggest routing changes, don't apply them.
heimdall-dream --repo . run --json

# also START one bounded, reversible experiment (safe per the keep-gate: unvalidated,
# snapshotted, auto-rolled-back next cycle if it does not beat baseline):
heimdall-dream --repo . run --overnight --start-experiment --json

Read the full file on GitHub · 151 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. 2d ago First seen · 151 lines · 0 tokens per session scan A 514575e6f807

Subscribe to this mod's changes

dream is a command published in the GitHub repository randomittin/heimdall (5 stars, last pushed 12d ago), licensed MIT. It adds 47 tokens to every session and 1,896 once invoked, about $0.0002 per session on Opus 5. 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 2026-08-31.