rlm

A method for processing documents or codebases that are too large to fit comfortably into one agent session, using smaller delegated analyses.

In plain words
What is it for?
Use it to analyze very long documents or extensive codebases through task files, background queries, progress checks, and a collected result.
Why use it?
It reduces the need to load an entire large context at once and organizes results from multiple depth-limited workers.

Skill for Claude CodeCodex

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 skills/dmoskov/sharedskills/rlm
Any agent
npx skills add dmoskov/sharedskills --skill rlm
Clone the repo
git clone --depth 1 https://github.com/dmoskov/sharedskills

Made for: Claude Code, Codex.

Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 980 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 $0.00049 $0.00980
Opus 5 $0.00024 $0.00490
Sonnet 5 $0.00010 $0.00196
Haiku 4.5 $0.00005 $0.00098

Measured 2d ago against content hash 252505aa28fa, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

rlm 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.

skills/rlm/SKILL.md · 80 lines

How it starts

The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Recursive Language Model (RLM)

You have access to rlm-query and rlm-batch for delegating long-context work to sub-agents. Do not try to orchestrate the work yourself — your job is to set up the workspace and immediately hand off to a depth-0 orchestrator.

Setup

SKILL_DIR="$(dirname "$(readlink -f ~/.claude/skills/rlm/SKILL.md 2>/dev/null || echo ~/.claude/skills/rlm/SKILL.md)")"
export PATH="${SKILL_DIR}/scripts:$PATH"

What to do

  1. Write a prompt file describing what needs to be done, including paths to any input files:

    TASK="my-task"
    mkdir -p .rlm/$TASK
    cat > .rlm/$TASK/task.md << 'EOF'
    Analyze /path/to/big-document.txt for liability risks...
    EOF
    
  2. Launch the root orchestrator in the background so you can give the user progress updates while it runs:

    rlm-query .rlm/$TASK/task.md .rlm/$TASK/result.out \
        --task $TASK --model opus --max-depth 2 &
    RLM_PID=$!
    
  3. Poll for progress and keep the user informed. Check active sub-agents, completed results, and whether the orchestrator has finished:

    # How many sub-agents are running?
    tmux list-sessions 2>/dev/null | grep -c "rlm-$TASK" || echo 0
    # How many results are in so far?
    ls .rlm/$TASK/results/*.out 2>/dev/null | wc -l
    # Is the orchestrator done?
    kill -0 $RLM_PID 2>/dev/null && echo "still running" || echo "done"
    

    Tell the user things like "3 sub-agents active, 5/16 chunks processed so far" while waiting.

  4. Read the result once the orchestrator finishes:

    wait $RLM_PID
    cat .rlm/$TASK/result.out
    

The orchestrator receives the full RLM instructions (rlm-agent.md) automatically and knows how to split, delegate to sub-agents, and aggregate. You don't need to read those instructions yourself.

Configuration

Set these via flags on rlm-query or environment variables. Choose based on the task:

  • --model / RLM_MODEL (default: opus) — Model for the orchestrator and sub-agents. Use opus for the root orchestrator (it needs to plan and decompose). Sub-agents doing straightforward work (classification, extraction) can use sonnet — the orchestrator can pass --model sonnet when it calls rlm-batch.

Read the full file on GitHub · 80 lines

Files

What ships with it

3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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 · 80 lines · 49 tokens per session scan A 252505aa28fa

Subscribe to this mod's changes

rlm is a skill published in the GitHub repository dmoskov/sharedskills (2 stars, last pushed 11d ago), licensed MIT. It adds 49 tokens to every session and 980 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.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

obra/superpowers · 37 tokens

chat-pet-sprite-creation

Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.

microsoft/vscode · 53 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

agent-host-chat-contributions

Build and review cross-cutting agent-host chat behavior through lifecycle contributions. Use when adding turn lifecycle side effects, prompt or context injection, restored-history transformation, protocol-action observation, or when reviewing changes that add code to AgentSideEffects or AgentService.

microsoft/vscode · 56 tokens

auto-perf-optimize

Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.

microsoft/vscode · 62 tokens