Skip to main content
plugin

perf — Claude Code plugin: performance analysis

This plugin runs a structured performance investigation with baselines, profiling, and hypothesis tests. Use it when a code path needs evidence before changes.

MIT ★ 2,003 updated 2026-07-26

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add composio-community/awesome-claude-plugins
/plugin install perf@awesome-claude-plugins
Open on GitHub

What perf does

The /perf command applies a sequential workflow to performance work. It records a baseline, runs one controlled experiment at a time, profiles the relevant code, and examines hotspots. Results are consolidated into one baseline for each version.

The plugin stores its output under {state-dir}/perf/, including an investigation record, Markdown investigation files, and versioned baseline data. Its internal skills gather hypotheses from Git history and evidence, discover code paths before profiling, test theories, analyze findings, and write structured logs. A hypotheses file can be supplied with --hypotheses-file <path> during the hypotheses phase.

The workflow is governed by docs/perf-requirements.md and docs/perf-research-methodology.md. Experiments run separately, so the order of the investigation is part of how its results are interpreted.

What it helps with

  • establish a performance baseline for a version
  • locate hotspots through code profiling
  • test performance hypotheses with controlled runs
  • support recommendations with collected evidence
  • store investigation logs and baseline files

Example prompts for the agent

  • Run /perf to investigate a slowdown in the current version.
  • Run /perf with hypotheses from hypotheses.md.
  • Use /perf to profile the relevant code paths and record the findings.

How to install perf

Plugin marketplace

/plugin marketplace add composio-community/awesome-claude-plugins
/plugin install perf@awesome-claude-plugins

Step by step, with commands for each platform: Claude Code plugins: marketplaces, installing, and building your own

Questions about perf

How do I install perf in Claude Code?

Add the marketplace with /plugin marketplace add composio-community/awesome-claude-plugins, then install the plugin with /plugin install perf@awesome-claude-plugins.

What does perf do?

It runs a staged performance investigation: baseline collection, controlled experiments, profiling, hotspot analysis, and evidence-backed recommendations.

Where does perf save its results?

The plugin writes investigation data and version baselines under {state-dir}/perf/, including investigation.json, investigation documents, and <version>.json files.

Source text

perf · MIT
# perf

Rigorous performance investigation workflow. `/perf` enforces sequential benchmarks, minimum run durations, and evidence-backed decision-making.

## Command

```
/perf
```

## What it does

- Establishes a baseline and persists results under `{state-dir}/perf/`
- Runs controlled experiments one at a time
- Performs profiling and hotspot analysis
- Consolidates findings into a single baseline per version

## Requirements

All behavior is governed by:
- `docs/perf-requirements.md`
- `docs/perf-research-methodology.md`

## Inputs

- Hypotheses can be supplied via `--hypotheses-file <path>` during the hypotheses phase.

## Skills

- `perf-theory-gatherer` - Hypothesis generation (git history + evidence)
- `perf-code-paths` - Code-path discovery before profiling
- `perf-theory-tester` - Controlled experiments for hypotheses
- `perf-analyzer` - Evidence-backed perf recommendations
- `perf-investigation-logger` - Structured log entries with evidence

## Artifacts

- `{state-dir}/perf/investigation.json`
- `{state-dir}/perf/investigations/<id>.md`
- `{state-dir}/perf/baselines/<version>.json`

© Community · MIT · composio-community/awesome-claude-plugins

Similar plugins

All plugins for DevOps and infrastructure →

From the same set68

The whole set →
developer-growth-analysis developer-growth-analysis analyzes recent Claude Code chat history to detect skill gaps, pulls learning resources from HackerNews, and delivers an individual growth report. ★ 2,003repo stars security-guidance security-guidance — a Claude Code hook flags possible security trouble while files are edited, including command injection, XSS, and risky code patterns. ★ 2,003repo stars documentation-generator documentation-generator — a Claude Code plugin that writes documentation from your source files: API reference, module descriptions, a project overview. Useful when the code exists and the docs don't. ★ 2,003repo stars audit-project audit-project — a Claude Code plugin running multi-agent iterative code review. Sub-agents scan, fix, and re-check until the issue list is empty. ★ 2,003repo stars mcp-builder mcp-builder — a Claude Code plugin that helps write MCP servers to connect LLMs with external services. Provides instructions on structure, data types and documentation patterns. ★ 2,003repo stars backend-architect Guides API design, database optimization, and scalable backend architecture. Activate it when designing server endpoints, securing authentication systems, or tuning query performance. ★ 2,003repo stars bug-fix bug-fix — a Claude Code plugin that opens a GitHub issue for a bug, branches off to fix it, and tests the change before it merges. ★ 2,003repo stars debugger The debugger plugin for Claude Code helps diagnose runtime errors, failing test suites, and unexpected software behavior during development. ★ 2,003repo stars

Project sponsors

Companies that keep this analytics open