Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Makio64/gpu-perf-agent --skill skillgit clone --depth 1 https://github.com/Makio64/gpu-perf-agentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/makio64/gpu-perf-agent/skill)<a href="https://agentmods.dev/skills/makio64/gpu-perf-agent/skill"><img src="https://agentmods.dev/badge/skills/makio64/gpu-perf-agent/skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/makio64/gpu-perf-agent/skill"><img src="https://agentmods.dev/badge/skills/makio64/gpu-perf-agent/skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00031 | $0.01142 |
| Opus 5.5 | $0.00012 | $0.00457 |
| Sonnet 5.5 | $0.00006 | $0.00228 |
| Haiku 4.5 | $0.00003 | $0.00114 |
Grade A, and why
gpu-perf-agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
GPU performance profiling
Requires Node.js 22+ and Chrome/Chromium. Use a project's installed gpu-perf-agent, npx gpu-perf-agent, or node src/cli.js from the tool checkout. The default CDP runner has no required npm dependencies. doctor --quick --json checks local setup; a full doctor --json also probes browser capabilities.
Efficient capture
npx gpu-perf-agent agent --url http://127.0.0.1:5173 --out reports/base.json
# After a change, repeat with the same settings:
npx gpu-perf-agent agent --url http://127.0.0.1:5173 --out reports/candidate.json
npx gpu-perf-agent agent --base reports/base.json --candidate reports/candidate.json
agent uses instrumentation, six 250 ms samples, and two warmups. It saves full evidence and prints a bounded digest with up to three warnings. --json prints structured compact output. Use the saved report for detailed recommendations and measurements. Missing evidence is not a zero measurement.
For repeated captures, reuse Chrome:
npx gpu-perf-agent serve --port 9099 --auto-instrument
npx gpu-perf-agent agent --server http://127.0.0.1:9099 --url http://127.0.0.1:5173 --out reports/candidate.json
Each job gets a fresh browser context by default. Use --context shared only for intentional stateful measurements. Configure browser launch flags on serve. Paths in HTTP jobs are resolved on the service machine. run --cdp 9222 also attaches to an existing debug-enabled Chrome; it creates profiling pages and leaves the browser running afterward.
Choosing measurement settings
--file page.html --file-root .serves local HTML and its project imports. A running dev server is preferable for bundled applications.--api webgpu,--api webgl2, or--api webglavoids unrelated API probes;autoprobes WebGPU and WebGL2.--wait-for-hookwaits for asynchronousglobalThis.__gpuReportBenchsetup.--wait-condition 'window.ready === true'supports custom readiness, bounded by--wait-condition-timeoutin milliseconds.- Keep fixed sampling for comparisons.
--adaptivemay stop after three samples when FPS CV is below 1.5%; it is a turnaround aid, not proof of statistical convergence or GPU stability. --preset confirmprovides 15 × 500 ms samples with four warmups. Useabwith--base-url,--candidate-url, and--rounds 4to interleave A/B runs in one browser. Local alternatives are--base-fileand--candidate-file.--traceadds Chrome diagnostics;--minimal-tracelimits categories to GPU events.--raw-traceretains the trace. Keep tracing settings identical between compared captures.--htmlwrites an interactive HTML report;--html-outsets its path.--screenshotsaves a PNG. Screenshot capture alone does not establish visual equivalence.--executable-pathorCHROME_PATHselects a browser. The optional--runner playwrightneeds Playwright installed separately.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · -97 lines · -46 tokens per session d6e8956bdf9f
- 1mo ago Changed · +35 lines 908de29d0183
- 1mo ago First seen · 118 lines · 77 tokens per session scan A 41e933431b01
gpu-perf-agent is a skill published in the GitHub repository Makio64/gpu-perf-agent (3 stars, last pushed 1mo ago), licensed MIT. It adds 31 tokens to every session and 1,142 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
debug-optimize-lcp
Guides debugging and optimizing Largest Contentful Paint (LCP) using Chrome DevTools MCP tools. Use this skill whenever the user asks about LCP performance, slow page loads, Core Web Vitals optimization, or wants to understand why their page's main content takes too long to appear. Also use when the user mentions…
systematic-debugging
Use when debugging a failing test, build error, or runtime issue that isn't immediately obvious. Guides a 4-phase root cause analysis instead of random fix attempts.
log-error-digest
Analyze log files to troubleshoot errors, identify peak error periods, and produce error clustering, frequency statistics, and time distribution reports. Supports JSON, syslog, and Nginx formats with automatic detection. Use when a user uploads a .log file and asks to analyze errors, find patterns, debug issues, or…
cwv-optimizer
Diagnose and fix Core Web Vitals issues on AEM Edge Delivery Services pages. Goes deeper than generic CWV advice by understanding EDS-specific performance patterns including the 100KB LCP budget, E-L-D loading phases, block rendering behavior, and third-party script impact. Produces specific fixes for LCP, CLS, and…
incident-response-incident-response
Coordinate production incident triage, mitigation, investigation, recovery, and follow-up. Use when responding to outages or other incidents that require structured operational coordination.
ci-fixer
CI failures - read error, minimal fix, verify.