Back to Skills
Document Processing

RLM Reader

Vetted and production-tested agent skill.

RLM Reader ā€” Vetted & Hardened by AgentVet

**šŸ›”ļø AgentVet Vetted ā€” Audited 2026-04-20**

**Overall Rating: LOW RISK**

**Audit Method: Cross-model verification (DeepSeek V4 Pro + Grok 4.3)**

---

Overview

Recursive Language Model (RLM) pattern for processing arbitrarily long documents and datasets. Implements the MIT RLM paper (arxiv 2512.24601) ā€” treats input as an external environment, writes code to search/slice it, and spawns sub-agents for parallel reading of specific regions.

When RAG summarization would destroy critical detail, RLM Reader preserves everything.

What We Found

SeverityFinding
**LOW**Subagent spawn paths not scope-bounded ā€” could read outside intended files
**LOW**No timeout enforcement on recursive reads

Patches Applied

  • **Fix 1:** Filesystem scope bounding ā€” subagents restricted to declared paths only
  • **Fix 2:** Per-read timeout enforcement with graceful degradation

Trust Badge

ā•”ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•—
ā•‘   šŸ›”ļø  AGENTVET VETTED                           ā•‘
ā•‘   RLM Reader v1.0 ā€” LOW RISK                    ā•‘
ā•‘   Audited 2026-04-20                             ā•‘
ā•‘   Method: Cross-model (DeepSeek V4 + Grok 4.3)   ā•‘
ā•‘   Patches: 2 minor fixes applied                 ā•‘
ā•šā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•ā•

---

*AgentVet provides independent security vetting. This skill is not affiliated with its original creators.*

Get the full skills pack

Enter your email to download all 15 AgentVet skills as a pack. No spam, just occasional updates on new skills.

AgentVet provides independent security vetting. Skills are not affiliated with their original creators unless stated.