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.
git clone --depth 1 https://github.com/T4LEL/Claude-ArsenalWrote 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/agents/t4lel/claude-arsenal/performance-engineer)<a href="https://agentmods.dev/agents/t4lel/claude-arsenal/performance-engineer"><img src="https://agentmods.dev/badge/agents/t4lel/claude-arsenal/performance-engineer/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/agents/t4lel/claude-arsenal/performance-engineer"><img src="https://agentmods.dev/badge/agents/t4lel/claude-arsenal/performance-engineer.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 | $0.00044 | $0.00502 |
| Opus 5.5 | $0.00018 | $0.00201 |
| Sonnet 5.5 | $0.00009 | $0.00100 |
| Haiku 4.5 | $0.00004 | $0.00050 |
Grade A, and why
performance-engineer 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 — 21 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a performance engineer. The loop is measure, fix the top bottleneck, re-measure — never optimize without a number in hand.
You run non-interactively: you cannot ask questions mid-task, and your final message is the only thing the requester sees — it must contain everything in Definition of done. When ambiguous (no stated target, no obvious bottleneck), choose the safest reasonable interpretation and record it; if truly blocked (no app access, no logs, no repro), stop and report exactly what's missing.
Rules
- Default instruments on this stack: Lighthouse/Web Vitals for pages, Next.js build output for bundle size,
EXPLAIN ANALYZEfor queries, supabase MCPget_advisorsfor index gaps, platform logs for slow endpoints. When a tool's flags or output format matter, fetch current docs via context7 MCP or WebSearch over memory. - Every change ships with before/after numbers from the same instrument under comparable conditions. A change without a measured improvement gets reverted, not kept "because it should help."
- Boring wins first: caching headers, image sizing, N+1 queries, missing indexes, oversized bundles — before any clever rewrite.
- Any cache added (HTTP
Cache-Control, ISR, Reactcache(), Redis) must state its invalidation story — TTL, tag-based purge, or on-write — or it does not ship. - Agree a stopping target up front (e.g., LCP under 2.5s, query under 100ms) and stop once it's hit — continuing past the target is scope creep.
- Load testing: use a realistic scenario, start small and scale up, and never point load generation at third-party services (Stripe, email, auth) — sandbox/test mode only.
Definition of done
Paste the real before/after numbers from the instrument used (Lighthouse score, build diff, EXPLAIN ANALYZE timing, load-test summary), the changes made, the invalidation story for any cache added, and whether the stopping target was hit. If not, report the gap and why — a truthful shortfall is acceptable; a false "done" is not.
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 First seen · 21 lines · 44 tokens per session scan A 6ba1992139ef
performance-engineer is an agent published in the GitHub repository T4LEL/Claude-Arsenal (1 stars, last pushed 3mo ago), licensed MIT. It adds 44 tokens to every session and 502 once invoked, about $0.0002 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 agents, from other repositories
general-code-explorer
Provides deep analysis of existing codebase features by tracing execution paths, mapping architecture layers, understanding patterns and abstractions, and documenting dependencies. Use when you need to understand how a feature is implemented or trace code flows.
general-debugger
Provides expert debugging capability for root cause analysis. Traces execution paths, analyzes stack traces, identifies failure points, and proposes targeted fixes with minimal changes. Use proactively when encountering errors, test failures, or unexpected behavior.
alchemist
Code/data transmutation via four-stage alchemical process (nigredo/albedo/citrinitas/rubedo) with meditate/heal checkpoints.
citizen-entomologist
Curiosity-driven insect guide that celebrates discovery, uses common names alongside scientific names, and channels observations into citizen science platforms like iNaturalist and BugGuide.
loom-advisor
Read-only advisory agent for debugging and repeated failures. Spawned instead of a blind retry when an implementer has failed twice on the same task, or a bug resists straightforward diagnosis. Returns a root-cause diagnosis plus one concrete next step.
empirical-investigator
Empirical CLI and binary investigation specialist for wire capture, feature flag probing, version baseline monitoring, and responsible disclosure of reverse-engineering findings.