Plugin

agent-trace-triage

Agent Trace Triage is a Claude Code plugin for developers building agentic AI systems (LangGraph, CrewAI, custom tool loops, MCP agents). When an agent run fails, stuck in a loop, wrong tool, or burning tokens. the user runs /agent-trace-triage with a JSON/JSONL trace file. The plugin: • Parses traces (JSONL, OpenAI tool_calls, partial LangSmith/OTEL-style exports) • Detects loops deterministically (identical repeat, ping-pong, retry-without-progress) via bundled Node scripts • Classifies failures into five buckets: tool selection, arguments, execution, orchestration, recovery policy • Writes a markdown triage report with targeted fix recommendations • Scaffolds regression tests (pytest / vitest templates) No MCP servers, no external API calls, no npm dependencies. Requires Node.js 18+ for local scripts only. MIT licensed.

• Published source
• Repository

Analysed

11 October 2026

Last observed

11 October 2026

What the code does

Static analysis found none of the capabilities we track, such as shell commands, network requests or file writes, in the latest version.

Having a capability is not wrongdoing; many tools need these to do their job. It tells you what to check before you install.

This is public metadata, not a safety certification or complete assessment. Static analysis cannot establish what a tool does at runtime.