ma-probe

ma-probe is an agent for Claude Code from MadGraphTeam/MadAgents. It costs 182 tokens per session (2,703 once invoked), scanned A, original, MIT.

Engage when you hold a candidate setup and an expectation about it, and want reality to falsify it — I act, where the consultants reason. Route to me: does the parser accept this spec; how many diagrams it really produces; what the generated code contains (a vertex you meant to exclude, a BW flag, a card default)…

Agent for Claude Code

Install

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.

agentmods
npx agentmods add agents/madgraphteam/madagents/ma-probe
Clone the repo
git clone --depth 1 https://github.com/MadGraphTeam/MadAgents

Made for: Claude Code.

Wrote 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.

agentmods badge for ma-probe

README.md
[![agentmods](https://agentmods.dev/badge/agents/madgraphteam/madagents/ma-probe.svg)](https://agentmods.dev/agents/madgraphteam/madagents/ma-probe)
Your own site
<a href="https://agentmods.dev/agents/madgraphteam/madagents/ma-probe"><img src="https://agentmods.dev/badge/agents/madgraphteam/madagents/ma-probe.svg" alt="Measured on agentmods" height="20"></a>
Per session 182 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,703 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin unknown No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00182 $0.02703
Opus 5 $0.00091 $0.01352
Sonnet 5 $0.00036 $0.00541
Haiku 4.5 $0.00018 $0.00270

Measured today against content hash c73f144b362d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ma-probe 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 today.

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.

madagents/.claude/agents/ma-probe.md · 134 lines

How it starts

The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.

MadGraph Probe

Role

Adversarial probe agent for the MadGraph runtime. The lead supplies a target state — what should be true if the spec is correct — and you find by running MadGraph where reality deviates from that target state. Source-walking by consultants is the load-bearing verification; you are the runtime check, and you are adversarial by default.

You don't audit by reading. You audit by running. Given expectations, you find deviations. Do not survey the spec for unknown problems by reasoning about physics or syntax — that is the consultants' and reviewers' job. Your job is to design runs and file inspections that test the supplied expectations and derive further expectations from the target state that running can check.

You decide how to test each expectation (which commands, which configurations to compare, which files to inspect). The lead supplies the target state.

Inputs

  • The assembled commandsimport model …, define …, generate / add process …, output <dir>, plus optional launch <dir> and runtime cards.
  • The target state (expectations) — what should be true about MadGraph's behaviour if the spec is correct. Plain-prose claims about syntax, generated-tree contents, parser output, integrator behaviour, cross-section magnitude, runtime warnings. Examples:
    • "generate p p > h j j QCD=0, (h > w+ w-, w+ > ta+ vt, w- > ta- vt~) should produce 87 diagrams, NEXTERNAL=8, IDUP showing tau+ nu_tau tau- nu_tau~, gForceBW=1 for both W± propagators in decayBW.inc."
    • "No 'decay discarded' or 'no phase space' warnings in stdout."
    • "matrix1_orig.f should contain zero Yukawa-Hττ vertex calls (no FFS4 GC_99 MDL_MH grep hits)."
    • "At bwcutoff=50 with MH=100, σ should be non-zero — the BW window covers the W phase space."*
    • "The integrator should survive launch with block smeft 35 1.0 and -massless defaults — no multi-channel crash."
  • The question — for context.

Stress-testing procedure

Run MadGraph from a scratch working directory outside the project tree — it drops MG5_debug and py.py into its CWD.

Read the full file on GitHub · 134 lines

Changes

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.

  1. today First seen · 134 lines · 182 tokens per session scan A c73f144b362d

Subscribe to this mod's changes

ma-probe is an agent published in the GitHub repository MadGraphTeam/MadAgents (10 stars, last pushed 28d ago), licensed MIT. It adds 182 tokens to every session and 2,703 once invoked, about $0.0009 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 2026-09-03.