performance-analysis

name	performance-analysis
description	Use for PcapRaven worst-case complexity review, bounded benchmark design, integer timing summaries, scalability analysis, and performance regression triage.

Performance Analysis

Procedure

1. Read AGENTS.md, docs/SECURITY_MODEL.md, docs/PERFORMANCE.md, and the active roadmap scope. Performance work must not weaken safety or semantics.
2. Enumerate major loops, maps, sorts, retained collections, parser revisits, output materialization, and attacker-controlled cardinalities. State a worst-case complexity and explicit finite bound for each.
3. Search for front removal/insertion, nested scans, repeated cloning, hidden whole-input buffering, recursive traversal, unchecked growth, and accidental floating-point use in exact analytical paths.
4. Use release builds, synthetic sanitized workloads, warmups, multiple odd sample counts, monotonic nanosecond clocks, and integer medians. Suppress result output without suppressing failures.
5. Include at least two controlled scales and report capture bytes, record counts, environment, toolchain, revision, minima, medians, maxima, and growth ratios. Never invent a threshold or measurement.
6. Reproduce regressions, profile the owning layer, and prefer the smallest complexity or allocation fix. Re-run exact behavior, schema, privacy, and full quality gates after any optimization.
7. Distinguish benchmark foundation, smoke execution, measured baseline, and accepted gate. Unrun acceptance work remains explicitly pending.