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 MadGraphTeam/MadAgents --skill train-docsgit clone --depth 1 https://github.com/MadGraphTeam/MadAgentsWrote 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/madgraphteam/madagents/train-docs)<a href="https://agentmods.dev/skills/madgraphteam/madagents/train-docs"><img src="https://agentmods.dev/badge/skills/madgraphteam/madagents/train-docs/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/madgraphteam/madagents/train-docs"><img src="https://agentmods.dev/badge/skills/madgraphteam/madagents/train-docs.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 | $0.00040 | $0.03323 |
| Opus 5.5 | $0.00016 | $0.01329 |
| Sonnet 5.5 | $0.00008 | $0.00665 |
| Haiku 4.5 | $0.00004 | $0.00332 |
Grade A, and why
train-docs 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 2mo 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 — 207 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Train Documentation
Run one cycle of the documentation improvement loop with parallel question processing.
Parameters
$ARGUMENTS
Parse key=value pairs from the input. All are optional — use defaults if not specified.
- questions: Comma-separated list of questions, or a path to a JSON file. If not given, generate questions.
- count: Number of questions to generate (default: 3). Ignored if questions are provided.
- focus: Topic focus for question generation. Example:
focus=NLO matching - requirements: Extra constraints for generation. Example:
requirements=must involve Pythia8
Workspace Setup
Answerers and verifiers must be spawned as teammates (via TeamCreate), NOT as subagents (via the Agent tool). Teammates run as independent Claude Code sessions that can dispatch their own subagents. Subagents cannot dispatch further subagents — so a verifier spawned as a subagent would be unable to delegate to madgraph-operator, script-operator, etc.
- Answerers:
agent_type: "orchestrator"— full MadAgents orchestrator. - Verifiers:
agent_type: "verifier"— MadAgents orchestrator with built-in verification workflow and restricted subagent roster.
Teammates start in /output (the lead's working directory). Always specify absolute paths (e.g., /output/train/q000/) when telling a teammate where to write its output.
Create ./train/ with a subdirectory per question: q000/, q001/, etc. Each question directory gets: answer/, verify/, grade/, diagnose/.
Workflow
Execute all steps in sequence without pausing for confirmation — the user will be consulted only where the skill explicitly says "Discuss with user."
1. Get questions
If questions were provided, use them. Otherwise invoke /generate-questions with the count, focus, and requirements parameters.
The generated questions include reference_answer fields. These are unverified and must not be visible to answerer subagents.
Write the full questions data (including reference_answer fields) to ./train/questions.json. This file will be hidden from answerers and re-answerers via hide_paths.
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.
- 2mo ago First seen · 207 lines · 40 tokens per session scan A b2e542d97d66
train-docs is a skill published in the GitHub repository MadGraphTeam/MadAgents (15 stars, last pushed 3mo ago), licensed MIT. It adds 40 tokens to every session and 3,323 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.
train-docs compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| train-docsMadGraphTeam/MadAgents | Skill | 15 | 3mo ago | MIT | A | 40 | Claude Code |
| imagegenopenai/codex | Skill | 133k | yesterday | Apache-2.0 | A | 113 | Codex |
| playwright-climicrosoft/playwright | Skill | 99k | yesterday | Apache-2.0 | A | 19 | Claude Code |
| local-ai-agentsmicrosoft/ai-agents-for-beginners | Skill | 79k | 4d ago | MIT | A | 200 | Claude Code, Codex |
| systematic-debuggingobra/superpowers | Skill | 313k | 3d ago | MIT | A | 21 | Claude Code |
| next-cache-components-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 95 | Claude Code |
| next-partial-prefetching-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 103 | Claude Code |