agent-platform-agent-tracing-standardslisted

# Standardize agent observability with tracing spans > OpenTelemetry-aligned span taxonomies for LLM calls, tool executions, and handoffs enabling real debugging. **Track:** 🤖 AI Engineering · **Domain:** Agent Platform Engineering · **Level:** advanced · **~40 min** **Who this is for:** AI Engineers, ML Engineers, LLM App Developers, Agent Builders ## When to Use This Skill OpenTelemetry-aligned span taxonomies for LLM calls, tool executions, and handoffs enabling real debugging. Use it whenever a matching task appears in conversation — the agent loads these instructions on demand. ## Steps 1. Define span hierarchy: session → plan → step → tool-call → sub-calls 2. Attach token counts, costs, model versions, and prompt hashes to every span 3. Propagate trace context across queues/services so retries stay linked 4. Sample intelligently: 100% of errors/tool-failures, fraction of successes 5. Build the 'why did it do that' query: input hash → exact prompts retrieved 6. Version span schema; dashboards break silently otherwise ## Common Pitfalls - Each framework's proprietary format trapping traces in silos - PII flowing into trace storage without scrubbing ## Commands **Install with skills CLI** ```bash npx skills add aniruddhaadak80/skills --skill agent-platform-agent-tracing-standards ``` **Install globally** ```bash npx skills add aniruddhaadak80/skills --skill agent-platform-agent-tracing-standards -g ``` --- Part of [aniruddhaadak80/skills](https://github.com/