[Feature] WFGY problem codes as tags in AgentOps traces for multi-agent debugging

[onestardao]

Hi, thanks for AgentOps and the MCP server. This kind of observability is essential when debugging complex agent runs.

I work on WFGY, an open semantic “firewall” and diagnostic layer for LLM pipelines, based on a 16-problem failure map for RAG and agents:

ProblemMap and docs:
https://github.com/onestardao/WFGY/blob/main/ProblemMap/README.md

The same taxonomy is now used or referenced in Harvard MIMS ToolUniverse, QCRI Multimodal RAG Survey and the Innsbruck Rankify project.

Idea

When reading traces in AgentOps, users often see a failure but have no compact vocabulary to describe it.

I propose an optional “WFGY problem tag” feature:

allow users to tag a trace or span with one or more WFGY problem codes (No.1–No.16),

surface these tags in the UI and export,

optionally include a link to the relevant ProblemMap doc for each tag.

This keeps AgentOps neutral, but gives teams a shared language for what went wrong.

If this aligns with your roadmap, I can help provide the mapping text and short descriptions for each problem.