mapletechlabs/maple/maple-agent-tracing

Maple agent tracing (router)

This skill only picks the right per-framework skill.

Step 1: Find every agent in the repo

Look for LLM and agent dependencies in every app and service: package.json, pyproject.toml, requirements*.txt, uv.lock, pom.xml, build.gradle*, *.csproj, go.mod. A repo can have more than one; handle each one.

Step 2: Install the matching skill and follow it

Pick the first row that matches the service's dependencies. Install the skill with:

npx skills add MapleTechLabs/maple/skills --skill <skill> -y

Then read the installed SKILL.md and follow it. If npx skills is unavailable, read the file directly from https://raw.githubusercontent.com/MapleTechLabs/maple/main/skills/<skill>/SKILL.md.

Order matters: a framework row wins over the provider-SDK row, because frameworks depend on provider SDKs and instrumenting both records every model call twice. Mastra and eve depend on ai and LangChain.js on openai too; match the framework first, and don't route a LangChain.js service to the provider-SDK skill.

Step 3: Hand-off

After the per-framework skill's own checks pass, tell the user:

• which services you instrumented and with which skill;
• that each conversation appears in Maple under Agent Sessions (https://app.maple.dev/agent-sessions, or app.eu.maple.dev for EU organizations) once one conversation has run;
• anything the per-framework skill said is not captured for that framework (for example, cost or message content), so the gap is expected rather than a bug.