# Profile Browsing

> Measure Bitsocial Web loading, navigation, or interaction performance and investigate observed bottlenecks.

- Skill: `bitsocialnet/profile-browsing-5` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add bitsocialnet/profile-browsing-5`
- Raw SKILL.md: https://api.skillmd.com/api/skills/bitsocialnet/profile-browsing-5/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: bitsocialnet (https://skillmd.com/u/bitsocialnet)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/bitsocialnet/profile-browsing-5

---


# Browsing performance

Define the affected route/interaction and the symptom or comparison to establish. Verify routes in the affected workspace and wait for the actual content or control to be ready.

Reuse a compatible server in this worktree. If one is needed, the task owner starts it in an owned terminal, records its process/session, and stops only that server afterward. A profiling child does not manage servers. Other tasks may have valid Vite processes.

Keep one browser active machine-wide through `./scripts/pw-session.sh`. Use the `playwright-cli` skill for session lifecycle and affected-flow coverage. Browser work and other heavy checks remain serialized. Profile a small flow directly; delegate a substantial independent route set to `profiler` only when useful, with a supplied URL, unique session name, criteria, and evidence to return. Wait for its browser cleanup before another browser task starts.

`about/src/lib/react-scan.ts` exposes the raw react-scan report; confirm its actual shape and availability before interpreting it. Other workspaces may not expose the helper.

Read [measurement guidance](references/measurement.md) for browser observers, document-versus-hash timing, and this checkout's React evidence. Capture only what resolves the performance question; do not add instrumentation or new app tooling to satisfy a reporting template.

Compare the same narrow flow before/after with equivalent throttle, viewport, content, and capture settings. Report URLs, methods, observed cost, evidence paths, and unavailable metrics. Separate symptoms from inferred causes; cheap rerenders alone do not justify an optimization. Close the exact session on every exit path and leave preexisting servers/profiles untouched.