For "what does the agent produce for library X" — which spans, attributes, metrics, or config knobs — use the OpenTelemetry Ecosystem Explorer, which fully maps the Java agent and exposes an agent-friendly surface (Markdown indexes and resolved JSON, no scraping). Do not answer from model memory.
name → id → the instrumentation's JSON data URL (the URL embeds the content <hash>).
WebFetch that JSON URL — one self-contained record (a few KB): resolved configurations,
telemetry (spans with spankind + typed attributes, and metrics), javaagenttargetversions, semanticconventions, and scope.
Version-specific or "what changed between releases": https://explorer.opentelemetry.io/agent/javaagent/versions.md lists versions and marks the latest; https://explorer.opentelemetry.io/data/javaagent/versions/<version>-index.json gives the id→hash map for a version — a differing hash for the same id across two versions means that instrumentation changed. Schema: https://explorer.opentelemetry.io/schemas/javaagent-instrumentation.schema.json; use /llms.txt for the agent-oriented index and /llms-full.txt for the full documentation.
Prefer this Explorer data over the raw ecosystem-registry YAML on GitHub: the Explorer applies upstream metadata corrections that the raw registry does not.