How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

performance-profiling - Skill

Imported from sickn33/agentic-awesome-skills (skills/performance-profiling/SKILL.md) via skills.sh. Install upstream with npx skills add sickn33/agentic-awesome-skills --skill performance-profiling. Copyright stays with the author.

Performance Profiling

Measure, analyze, optimize - in that order.

🔧 Runtime Scripts

Inspect this bundled helper as code before use and run only against an authorized URL. Lighthouse launches a browser and can load third-party resources; lab scores do not establish field Core Web Vitals. The helper reports command/JSON failures and unavailable scores separately; inspect the returned error and do not treat an unavailable category as a zero score.

Script	Purpose	Usage
`scripts/lighthouse_audit.py`	Lighthouse performance audit	`python scripts/lighthouse_audit.py https://example.com`

1. Core Web Vitals

Targets

Metric	Good	Poor	Measures
LCP	< 2.5s	> 4.0s	Loading
INP	< 200ms	> 500ms	Interactivity
CLS	< 0.1	> 0.25	Stability

When to Measure

Stage	Tool
Development	Local Lighthouse
CI/CD	Lighthouse CI
Production	RUM (Real User Monitoring)

2. Profiling Workflow

The 4-Step Process

1. BASELINE → Measure current state
2. IDENTIFY → Find the bottleneck
3. FIX → Make targeted change
4. VALIDATE → Confirm improvement

Profiling Tool Selection

Problem	Tool
Page load	Lighthouse
Bundle size	Bundle analyzer
Runtime	DevTools Performance
Memory	DevTools Memory
Network	DevTools Network

3. Bundle Analysis

What to Look For

Issue	Indicator
Large dependencies	Top of bundle
Duplicate code	Multiple chunks
Unused code	Low coverage
Missing splits	Single large chunk

Optimization Actions

Finding	Action
Big library	Import specific modules
Duplicate deps	Dedupe, update versions
Route in main	Code split
Unused exports	Tree shake

4. Runtime Profiling

Performance Tab Analysis

Pattern	Meaning
Long tasks (>50ms)	UI blocking
Many small tasks	Possible batching opportunity
Layout/paint	Rendering bottleneck
Script	JavaScript execution

Memory Tab Analysis

Pattern	Meaning
Growing heap	Possible leak
Large retained	Check references
Detached DOM	Not cleaned up

5. Common Bottlenecks

By Symptom

Symptom	Likely Cause
Slow initial load	Large JS, render blocking
Slow interactions	Heavy event handlers
Jank during scroll	Layout thrashing
Growing memory	Leaks, retained refs

6. Quick Win Priorities

Priority	Action	Impact
1	Enable compression	High
2	Lazy load offscreen images; keep the LCP image eager	Workload-dependent
3	Code split routes	High
4	Cache static assets	Medium
5	Optimize images	Medium

7. Anti-Patterns

❌ Don't	✅ Do
Guess at problems	Profile first
Micro-optimize	Fix biggest issue
Optimize early	Optimize when needed
Ignore real users	Use RUM data

Remember: The fastest code is code that doesn't run. Remove before optimizing.

When to Use

This skill is applicable to execute the workflow or actions described in the overview.

Limitations

Use this skill only when the task clearly matches the scope described above.
Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.

Install performance-profiling and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

performance-profiling

About