MVPavan/via / .claude/skills/performance-optimization
performance-optimization
VerifiedUse for measured latency, throughput, memory or cost problems, profiling, or explicit resource targets. Skip prose optimization and speculative tuning.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.3 KB |
| agents/openai.yaml | config | 42 B |
| references/frontend-performance.md | markdown | 6.1 KB |
| references/measurement-recipes.md | markdown | 5.0 KB |
SKILL.md
View on GitHub---
name: performance-optimization
description: Use for measured latency, throughput, memory or cost problems, profiling, or explicit resource targets. Skip prose optimization and speculative tuning.
---
# Performance Optimization
Improve a named user outcome with comparable evidence. Prose shortening and
ordinary cleanup are outside this skill. For a newly introduced regression,
use systematic-debugging when causal localization is the main question.
1. Define the relevant metric, workload and target. If unclear, propose the
measurement or ask about the consequential tradeoff before tuning code.
2. Capture a baseline with representative data, cache state, concurrency and
environment. Repeat enough to estimate variability for the decision; use
percentiles when the service target concerns tail latency. Record the command,
conditions and results rather than assuming a fixed repeat count proves significance.
3. Locate the material cost using existing measurements, focused instrumentation
or profiling. Code inspection can suggest candidates; measurements determine
whether they matter. Use `references/measurement-recipes.md` for unfamiliar
tools and `references/frontend-performance.md` for browser performance.
4. Change one attributable factor where practical. For an inseparable set, explain
the coupling. Preserve required semantics, freshness, safety and accessibility.
5. Compare under the same conditions and run applicable correctness checks. Keep
changes whose benefit justifies their complexity; remove only your ineffective
experimental edits, preserving other work. A result within noise is inconclusive,
not evidence of a speedup. Record cost/latency/memory tradeoffs explicitly.
Keep a compact attempt record with baseline → result, decision and reason so failed
experiments are not repeated after context recovery. Persist it in the task/report
when the work needs handoff, not as an extra artifact for every tiny experiment.
For a material recurring risk, add an economical regression guard: query counts,
bounded allocations or a controlled benchmark, using existing tooling. Scale its
cost to the risk; avoid flaky wall-clock assertions and unnecessary frameworks.
Report measured outcomes, applicable checks and limits. Performance improvement
does not excuse a correctness regression or authorize unrelated edits.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |