https://developer.android.com/agents/skills/profilers/perfetto-sql/skill

To write valid Perfetto trace analysis SQL, you must use the Perfetto Standard Library schemas for every referenced table or view, enforce idempotent CREATE OR REPLACE patterns, and execute validated queries via trace_processor.

Handling incomplete slices and interval overlaps in PerfettoSQL requires checking for dur = -1 and using safe overlap logic with standards library features like SPAN_JOIN and intervals.overlap precedence.

PerfettoSQL query failures during argument extraction or thread joins often occur from unsafe assumptions; use EXTRACT_ARG for arguments, ensure correct join keys like utid/upid, and use GLOB instead of LIKE for string matching.

To safely re-run PerfettoSQL queries without manual drops, apply idempotent object creation rules using CREATE OR REPLACE and implement safe drop patterns for virtual tables to prevent re-run failures.

Yes, you can use trace_processor to validate schema and module usage before executing PerfettoSQL by running queries via the trace_processor wrapper with --query-string and validating against Perfetto Standard Library docs.

Perfetto SQL validation limitations involve trace boundary handling across typical trace time windows; queries must account for incomplete slices and avoid brittle logic by adhering to trace-specific safety rules.