Borrowing it
Nothing to install: this file belongs to rafaelsztutman/dtc-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/rafaelsztutman/dtc-mcp/main/.claude/commands/bench.mdgit clone --depth 1 https://github.com/rafaelsztutman/dtc-mcpWrote 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/commands/rafaelsztutman/dtc-mcp/bench)<a href="https://agentmods.dev/commands/rafaelsztutman/dtc-mcp/bench"><img src="https://agentmods.dev/badge/commands/rafaelsztutman/dtc-mcp/bench/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/commands/rafaelsztutman/dtc-mcp/bench"><img src="https://agentmods.dev/badge/commands/rafaelsztutman/dtc-mcp/bench.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 | $0.02115 |
| Opus 5 | $0.00015 | $0.01058 |
| Sonnet 5 | $0.00006 | $0.00423 |
| Haiku 4.5 | $0.00003 | $0.00212 |
Grade A, and why
bench 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 8d 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 — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/bench — head-to-head MCP benchmark driver
You are about to execute one batch of the dtc-mcp vs Klaviyo-MCP benchmark. The benchmark is organized along the conversation-length axis: 1-turn baseline, 2-turn short, 5-turn medium, 10-turn long. The hypothesis is that MCPs not designed for conversation continuity pay a compounding cost as turns grow, so dtc-mcp's stateful sandbox should outperform tool-list MCPs more sharply at longer conversations.
Execution model: every cell — single-turn or multi-turn — is run via
tsx bench/runner/cli.ts multiturn --cell <cellId>. That command spawns
ONE claude -p --input-format stream-json process, feeds it the cell's
prompts over stdin, parses the streamed output for per-turn responses +
aggregate usage, then records the result. One claude process = one MCP
transport, so dtc-mcp's per-connection sandbox state persists across the
whole trajectory (the empirically-verified property that motivates the
whole benchmark).
Grading is deferred to a separate LLM-as-judge phase after recording — there is no per-cell grading during a batch run.
Args
--batch <A|B|C|D>(default: A) — which batch to run- A = baseline (1 turn) × 3 tasks × 2 MCPs × 2 trials = 12 cells
- B = short (2 turns) × 2 tasks × 2 MCPs × 2 trials = 8 cells
- C = medium (5 turns) × 2 tasks × 2 MCPs × 2 trials = 8 cells
- D = long (10 turns) × 2 tasks × 2 MCPs × 2 trials = 8 cells
--retry <cellId>— re-run one failed/invalid cell, ignore others--limit N— only run the first N pending cells (useful for smoke tests)
How to execute
-
Verify the run is initialized. If
bench/results/has no recentstate.json, instruct the user to runtsx bench/runner/cli.ts initfirst and stop. -
Verify both MCPs are configured at the user level (
~/.claude.jsonmcpServers) so the spawnedclaude -pchild process sees them. There should be bothdtc-mcp(or whatever they named it) andklaviyo(or similar). If not, instruct the user to add the missing one and stop.
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.
- 8d ago First seen · 170 lines · 31 tokens per session scan A 6c7d2ca55924
bench is a command published in the GitHub repository rafaelsztutman/dtc-mcp (0 stars, last pushed 3mo ago), licensed MIT. It adds 31 tokens to every session and 2,115 once invoked, about $0.0002 per session on Opus 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 commands, from other repositories
verify
Run repository verification using the verification-loop skill.
qa-changes
This skill should be used when the user asks to "QA a pull request", "test PR changes", "verify a PR works", "functionally test changes", or when an automated workflow triggers QA validation of code changes. Provides a structured methodology for setting up the environment, exercising changed behavior, and reporting…
test-coverage
Analyze test coverage and identify the highest-value gaps to fill.
tdd
A command that follows test-driven development (TDD), a method where you write tests before the code they check. It moves through writing a failing test, adding the smallest implementation, and then improving the code.
check-dev
Type-check a Z specification with fuzz.
laravel-playwright
E2E Playwright patterns; use the laravel:e2e-playwright skill exactly as written.