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 agentmods add instructions/sanjoy1234/gauntlex/agents-mdgit clone --depth 1 https://github.com/sanjoy1234/gauntlexWrote 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/instructions/sanjoy1234/gauntlex/agents-md)<a href="https://agentmods.dev/instructions/sanjoy1234/gauntlex/agents-md"><img src="https://agentmods.dev/badge/instructions/sanjoy1234/gauntlex/agents-md.svg" alt="Measured on agentmods" 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 | $0.01011 | $0.01011 |
| Opus 5 | $0.00505 | $0.00505 |
| Sonnet 5 | $0.00202 | $0.00202 |
| Haiku 4.5 | $0.00101 | $0.00101 |
Grade A, and why
gauntlex AGENTS.md 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 4d 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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Instructions for AI coding agents (Codex, Cursor, Cline, Windsurf, Gemini CLI, Claude Code, and any other agent that reads this file) working in this repo.
What this project is
GAUNTLEX is an adversarial co-generation engine. It runs two agents concurrently against the same specification — Builder (writes the implementation) and Breaker (writes adversarial attacks against that same spec) — then an Arbiter scores every attack and produces an Adversarial Resilience Score (ARS). See README.md for the full pitch.
Package name on PyPI: gauntlex-ai. CLI entry point: gauntlex.
Setup
python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
gauntlex doctor # verify model connectivity + ChromaDB before anything else
Running the test suite
python -m pytest -q
Run this after any change under src/gauntlex/. The suite is fast (~2s,
500+ tests) because it mocks external model calls — it verifies wiring and
logic, not live model behavior. If you touch anything that calls out to a
real model, MCP transport, or CLI subprocess, also do a live smoke test (see
below) since the mocked suite will not catch integration-only breakage.
Live smoke test (do this for any change to cli.py, harness/, or mcp/)
gauntlex run --issue examples/demo_issue.md --mode quick --pretty
gauntlex status
gauntlex findings
--mode quick keeps this under ~2 minutes. Do not use --mode thorough for
routine verification.
Key CLI commands an agent will typically need
| Command | Purpose |
|---|---|
gauntlex run --issue <file|url> --mode quick|standard|thorough |
Run Builder+Breaker, get an ARS score |
gauntlex doctor |
Environment health check — run this first if anything fails |
gauntlex validate |
Dry run, zero attacks fired, just checks config/connectivity |
gauntlex findings |
Vulnerability-first summary of the last run |
gauntlex status |
List recent runs and pass/fail gate state |
gauntlex mcp-server |
Start GAUNTLEX as an MCP server (stdio) |
gauntlex integrate --platform all |
Wire GAUNTLEX's MCP config into every supported IDE/agent at once |
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.
- 4d ago First seen · 91 lines · 1,011 tokens per session scan A a56940ad279f
gauntlex AGENTS.md is an instructions file published in the GitHub repository sanjoy1234/gauntlex (1 stars, last pushed 1mo ago), licensed MIT. It adds 1,011 tokens to every session, about $0.0051 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 instructions, from other repositories
ai-appsec AGENTS.md
Instructions for subodhkc/ai-appsec, covering agents.md — haiec agent security mcp (phase -1 → phase 18), 1. workspace layout, 2. architectural principles (apply to every phase), 3. hard constraints (never violate) and 4. phase discipline.
skilltotal CLAUDE.md
Claude Code instructions for pezhik/skilltotal, covering claude.md, project, commands, install (editable) + dev deps and run one test file / one test.
MCP-Sentinel AGENTS.md
AGENTS.md instructions for BashaarJavaid/MCP-Sentinel, covering agents.md, project, where things live, conventions and commands.
arcjet-js AGENTS.md
AGENTS.md instructions for arcjet/arcjet-js, covering agent guidance, examples live in arcjet/examples, agent skills and integration work: review before a pr.
enforra AGENTS.md
Instructions for enforra/enforra, covering agents.md, enforra engineering doctrine, package boundaries, required workflow before pr and diagnostics.
promptsonar AGENTS.md
Instructions for meghal86/promptsonar, covering promptsonar agent instructions, operating principles, repo layout, common commands and change workflow.