context-labscontext-labsOfficialยท1 Agent Skills Included

HALO

Diagnose and fix AI agent failures using production traces

Analyzes execution traces from AI agents to find common failure modes, bottlenecks, and wasted tokens. Turns raw OpenTelemetry trace files into ranked diagnostic reports with concrete evidence and trace citations. Feeds findings into coding agents like Claude Code or Cursor so fixes ship faster with less manual debugging.
npx skills add context-labs/HALO --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install HALO?โ–ผ

Run `npx skills add context-labs/HALO --all -g -y` in your terminal to install all tools in this suite globally.

What does HALO do for AI agents?โ–ผ

HALO reads your agent's execution traces and identifies recurring failure patterns, slow steps, and costly tool calls. It produces a report you can hand to a coding agent to apply targeted fixes.

How do I debug my AI agent with HALO?โ–ผ

Capture OpenTelemetry-compatible JSONL traces from your agent, then run the HALO CLI with a diagnostic question. HALO explores the traces and returns evidence-backed findings with trace IDs.

Does HALO work with Claude Code and Cursor?โ–ผ

Yes. HALO's diagnostic reports are designed to be sent directly to coding agents like Claude Code, Cursor, or Codex, which then implement the recommended harness fixes.

What trace formats does HALO support?โ–ผ

HALO works with plain JSONL files containing OpenTelemetry-shaped spans, and can import traces from tools like Langfuse, Arize, or the OpenAI Agents SDK.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’