What if data is insufficient to make a recommendation?

Perf-analyzer will state that data is missing and request a re-run with additional logs or measurements.

Can I cite sources outside logs or code?

No. Only evidence that exists in logs or code may be cited, per the canonical contract.

How should I use next_steps in the output?

Use next_steps to decide whether to continue experiments, refine hypotheses, or stop and implement changes.

perf-analyzer

Scanned

npx machina-cli add skill ComposioHQ/awesome-claude-plugins/analyzer --openclaw

Files (1)

SKILL.md

779 B

perf-analyzer

Synthesize performance investigation results into clear recommendations.

Follow docs/perf-requirements.md as the canonical contract.

Inputs

Baseline data
Experiment results
Profiling evidence
Hypotheses tested
Breaking point results

Output Format

summary: <2-3 sentences>
recommendations:
  - <actionable recommendation 1>
  - <actionable recommendation 2>
abandoned:
  - <hypothesis or experiment that failed>
next_steps:
  - <if user should continue or stop>

Constraints

Only cite evidence that exists in logs or code.
If data is insufficient, say so and request a re-run.

Source

git clone https://github.com/ComposioHQ/awesome-claude-plugins/blob/master/perf/skills/analyzer/SKILL.mdView on GitHub

Overview

perf-analyzer synthesizes baseline data, experiment results, profiling evidence, and hypotheses into concise, evidence-backed recommendations. It follows the canonical contract in docs/perf-requirements.md to ensure consistency and traceability. If data is insufficient, it flags gaps and requests a re-run.

How This Skill Works

Provide inputs (baseline data, experiment results, profiling evidence, hypotheses tested, and breaking point results). The tool outputs a structured report with summary, actionable recommendations, any abandoned hypotheses, and next steps, strictly citing evidence from logs or code.

When to Use It

Diagnosing a performance regression after a release
Evaluating proposed optimizations to verify impact
Validating hypotheses against logs and profiling data
Benchmarking a new feature against baseline
Presenting findings to stakeholders for evidence-based decisions

Quick Start

Step 1: Gather baseline data, experiment results, profiling evidence, hypotheses, and breaking point results.
Step 2: Run perf-analyzer to produce a summary, recommendations, abandoned hypotheses, and next steps.
Step 3: Review the output and iterate with a re-run if data gaps are detected.

Best Practices

Cite only evidence that exists in logs or code
Align findings with docs/perf-requirements.md canonical contract
Predefine success criteria and hypotheses before analysis
Document assumptions and data gaps clearly
Tag and isolate which experiments support or refute each recommendation

Example Use Cases

Latency regression after release: baseline vs experiment showed 25% slower response times; recommendations target critical hot paths.
GC pauses identified: profiling showed long pause times; advised heap tuning and allocator changes.
Memory growth observed under load: suspected fragmentation; recommended refactor and pooling strategy.
Throughput improvement with feature flag: experiments confirmed higher throughput in optimized path.
I/O wait spike due to DB contention: recommended indexing and query plan improvements.

Frequently Asked Questions

Add this skill to your agents