dream

dream is a command for Claude Code from conorbronsdon/agent-context-os. It costs 17 tokens per session (2,031 once invoked), scanned A, original, MIT.

A command that runs a curator pass over a validated memory directory and creates a proposal for changes. Its available passes review memory content, merges overlaps, splits mixed files, or checks index and file agreement.

In plain words
What is it for?
Use it to review whether memories are still true, consolidate duplicates, divide multi-topic files, or find mismatches and misplaced files. Several listed review modes are not yet built.
Why use it?
It helps keep a collection of stored notes accurate, focused, and organised instead of requiring a manual review of every file.

Command for Claude Code

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/conorbronsdon/agent-context-os/dream
Clone the repo
git clone --depth 1 https://github.com/conorbronsdon/agent-context-os

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 dream

README.md
[![agentmods](https://agentmods.dev/badge/commands/conorbronsdon/agent-context-os/dream.svg)](https://agentmods.dev/commands/conorbronsdon/agent-context-os/dream)
Your own site
<a href="https://agentmods.dev/commands/conorbronsdon/agent-context-os/dream"><img src="https://agentmods.dev/badge/commands/conorbronsdon/agent-context-os/dream.svg" alt="Measured on agentmods" height="20"></a>
Per session 17 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,031 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 $0.00017 $0.02031
Opus 5 $0.00009 $0.01015
Sonnet 5 $0.00003 $0.00406
Haiku 4.5 $0.00002 $0.00203

Measured 4d ago against content hash c915f7db2bbe, 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 4d 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.

.claude/commands/dream.md · 157 lines

How it starts

The opening of the file, as written. The whole thing — 157 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/dream — on-demand memory curator pass

Substrate background: docs/dream-architecture.md. Curator prompts: scripts/dream/prompts/.

Usage

/dream             # default curator: rot
/dream rot         # content class: memory vs. world — "is it still true?"
/dream merge       # structural class: consolidate overlapping memories + collapse index lines
/dream split       # structural class: divide multi-concern files into focused ones
/dream lint        # structural class: index/file agreement, duplicates, misfiles, half-finished archives
/dream pattern     # NOT YET BUILT (v0.4)
/dream contradiction  # NOT YET BUILT (v0.5)
/dream untapped    # NOT YET BUILT (v0.6)
/dream audit       # NOT YET BUILT (v0.7)

Steps

1. Resolve curator name

If $ARGUMENTS is empty or rot, proceed with rot. If anything else, check whether scripts/dream/prompts/{name}.md exists. If not, list available curators and stop.

2. Resolve and validate the explicit memory dir

Run the executable validator before reading or writing memory state:

python3 scripts/dream/validate-memory.py resolve

Parse the returned JSON. Set MEMORY_DIR from memory_dir and treat repository_identity as the stable binding for this repository. The helper verifies .context-os/memory-directory, rejects non-canonical or symlinked paths, requires MEMORY.md and .context-os-repository, binds the marker to Git's resolved common directory identity (safe across linked worktrees), requires the memory dir to be its own git repo, refuses any memory remote, and requires tracked, staged, and untracked memory state to be clean before the pass. If host auto-memory left changes, stop so the user can review or snapshot them separately.

Only after that succeeds, generate TS in exact YYYY-MM-DDTHH-MM-SSZ form and preflight the target artifact path:

python3 scripts/dream/validate-memory.py artifact "$TS" --for-create

Create only the validated $MEMORY_DIR/.dreams/$TS path returned by the helper, then continue.

Read the full file on GitHub · 157 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. 4d ago First seen · 157 lines · 17 tokens per session scan A c915f7db2bbe

Subscribe to this mod's changes

dream is a command published in the GitHub repository conorbronsdon/agent-context-os (21 stars, last pushed today), licensed MIT. It adds 17 tokens to every session and 2,031 once invoked, about $0.0001 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 2026-08-30.