Performance Checklist: Agent Skill for Perf-Sensitive Reviews

Where this fits

Imagine you’re reviewing a PR that switches a scraping pipeline to unbounded concurrency, or one that tweaks staleTime on a React Query client. Without a systematic checklist, you might miss a subtle blocking call on the Tokio runtime or an N+1 query disguised as a refactor. This skill drops a focused performance review checklist into your agent’s context, so it can flag concrete regressions instead of guessing.

Why agents benefit

• Standardized review discipline: The checklist enforces consistent checks for async runtime, hot paths, renderer tuning, and token efficiency across any codebase.
• Severity bias guidance: Defaults to MEDIUM severity unless there’s a concrete hot-path regression, helping agents avoid crying wolf.
• Actionable specifics: Instead of generic “optimize performance,” it checks for spawn_blocking usage, query-client tuning, pre-measured layouts, and batch sizing — all things an agent can verify.
• Easy integration: One npx skills add command makes it available, with a documented authoritative source (docs/knowledge/performance-rules.md) for deeper context.
• Token-aware: Explicitly calls out minimizing prompt/context size, which keeps agent costs in check during long review sessions.

Practical scenarios

Reviewing a scraping pipeline change

Agent loads the skill and sees the hot-path checklist: bounded concurrency, reused page lifecycle, per-board rate limits. It flags a new tokio::spawn doing blocking I/O and suggests spawn_blocking.

Optimizing a React Query client for desktop

The skill notes “no refetch-on-focus” for desktop and reminds to tune staleTime/gcTime. Your agent catches a default setting that causes unnecessary network calls every focus.

Evaluating AI embedding token usage

The checklist’s token-efficiency section prompts the agent to audit batch sizes and model choice. It finds the code uses a larger model than needed for a fallback path, and recommends a cheaper alternative.

Add it to your agent workflow

Install with:

npx skills add saeedkolivand/ai-job-hunter-app --skill performance-checklist

Then instruct your agent (e.g., in a review prompt) to load the skill before analyzing perf-sensitive files:

Load performance-checklist skill before reviewing PR #123, focusing on async paths and renderer tuning.

The skill will pull the latest checklist from the README and apply it to your review.

Compare before adopting

Before committing, compare this skill against alternatives: check for recent commits (quality signal), star trend (community momentum), and whether it covers your stack (e.g., tokio vs. Node.js). Also read the README’s authoritative source — a skill that references a living doc is fresher than one that hardcodes rules.

Why it is worth tracking

With 51 stars and a clear focus on real-world perf bottlenecks, this skill is a solid addition to any agent’s toolkit. Watch for updates that extend coverage to other runtimes or frameworks. If you frequently review async-heavy or data-intensive code, evaluate it now — it might become your default perf gate. For more skills like this, browse the OpenAgentSkill catalog at https://www.openagentskill.com/skills/saeedkolivand-performance-checklist.