Six Reliability Primitives for LLM Agents: Software Artifact Package

An artifact package for the preprint Six Reliability Primitives for LLM Agents, published on May 12, 2026. It documents six small reliability libraries for tool-using LLM agents: AgentFit, AgentGuard, AgentSnap, AgentVet, AgentCast, and AgentBudget. The package includes the paper, metadata, and submission materials describing the artifact pattern across TypeScript, Python, and Model Context Protocol variants.

Use Cases

• Implementing reliability checks for LLM agents based on the described AgentFit, AgentGuard, AgentSnap, AgentVet, AgentCast, and AgentBudget libraries.
• Studying software artifact patterns for AI safety based on the included paper and submission materials.
• Comparing reliability library implementations across TypeScript, Python, and Model Context Protocol variants as documented in the package.

Strengths

• Includes six distinct reliability libraries (AgentFit, AgentGuard, AgentSnap, AgentVet, AgentCast, AgentBudget) for a multi-faceted approach.
• Provides cross-language documentation covering TypeScript, Python, and Model Context Protocol variants.
• Associated with a specific, cited academic preprint (DOI: 10.5281/zenodo.20074702).

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Harvard Dataverse

Collection Method

Artifact package created for an academic preprint.

Freshness

Last updated 2026-05-12 17:11:29; freshness should be verified.