Run agentic anomaly investigations with PyOD
Use PyOD's agent skill or MCP server to turn natural-language anomaly detection requests into repeatable detector selection, fitting, scoring, and investigation workflows.
npx skills add agentskillexchange/skills --skill run-agentic-anomaly-investigations-with-pyod
pip install pyod. For agent skill use, run pyod install skill or pyod install skill --project. For MCP clients, install pip install pyod[mcp] and start the server with pyod mcp serve.Use PyOD when an agent needs to investigate outliers or anomalies in tabular, time-series, graph, text, or image data and return a reproducible detection path. The operator installs PyOD’s agent skill or starts its MCP server, then asks the agent to choose detectors, run ADEngine-backed workflows, score records, and explain the anomaly evidence. The workflow input is a dataset plus an anomaly question; the expected outputs are detector choices, scores, plots or ranked records, and a reviewable explanation of why cases were flagged. This is not a generic Python library listing: the scope boundary is PyOD 3’s documented agentic anomaly detection activation paths, including `pyod install skill` for Claude/Codex-style skill use and `pyod mcp serve` for MCP-compatible clients.
What this skill actually does
Inputs and prerequisites: Python, pip, PyOD, optional MCP-compatible client.
Setup notes: Install with `pip install pyod`. For agent skill use, run `pyod install skill` or `pyod install skill –project`. For MCP clients, install `pip install pyod[mcp]` and start the server with `pyod mcp serve`.
Source and verification boundary: use http://pyod.readthedocs.io as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.