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/frontier-infra/roundtable/agents-mdgit clone --depth 1 https://github.com/frontier-infra/roundtableWhat 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 | $0.01271 | $0.01271 |
| Opus 5 | $0.00635 | $0.00635 |
| Sonnet 5 | $0.00254 | $0.00254 |
| Haiku 4.5 | $0.00127 | $0.00127 |
Grade E, and why
roundtable AGENTS.md scanned grade E with 4 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 yesterday.
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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
Keys live in `~/.config/roundtable/config.env` (dotenv, `chmod 600`, sourced — no parser). Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -fsSL https://roundtable.sh/install.sh | bash Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
- **Codex CLI** — `~/.codex/config.toml` `[mcp_servers.roundtable]`. Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -fsSL https://roundtable.sh/install.sh | bash How it starts
The opening of the file, as written. The whole thing — 137 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Roundtable
A guide for coding agents (and the humans wiring them up). Roundtable is a CLI + MCP server that convenes a council of frontier models on a question and, optionally, runs a chaired deliberation to consensus. It's a Frontier Infra project — we eat our own cooking: our own agents call Roundtable before the decisions that matter.
What it is
One command, several model "heads", run in parallel:
- Advisory (default) — every configured head answers once, blind to the others.
- Deliberation (
--rounds N) — heads read each other and revise; a Claude chair declaresCONSENSUS/CONTINUEand the run stops early on consensus. - Research (
--research) — web on, multi-step, for answers that need current facts.
Heads: grok · codex · openai · glm · minimax · claude · gemini. A head with
no key/CLI is skipped inline — the council runs on whatever subset is configured (so it
works on API keys alone, no local CLIs required).
Install
# Canonical one-liner
curl -fsSL https://roundtable.sh/install.sh | bash
# Or pip / uv (thin launcher around the same engine)
pip install roundtable
uv tool install roundtable
# Or Homebrew
brew install frontier-infra/tap/roundtable
# Or from a clone
git clone https://github.com/frontier-infra/roundtable && cd roundtable && ./install.sh --from .
The installer drops the engine under ${XDG_DATA_HOME:-~/.local/share}/roundtable and
symlinks roundtable into ~/.local/bin. If that's not on your PATH, the installer
prints the exact export PATH=... line to add.
Configure keys
roundtable auth # interactive: masked paste per provider, shows ✓/✗
roundtable auth anthropic # configure just one provider
roundtable doctor # which heads are configured + a reachability ping
Keys live in ~/.config/roundtable/config.env (dotenv, chmod 600, sourced — no parser).
You can also write it directly:
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.
- yesterday First seen · 137 lines · 1,271 tokens per session scan E e1684f7ca053
roundtable AGENTS.md is an instructions file published in the GitHub repository frontier-infra/roundtable (0 stars, last pushed 2mo ago), licensed MIT. It adds 1,271 tokens to every session, about $0.0064 per session on Opus 5. A static security scan graded it E with 4 findings (asks for root, downloads and executes remote code, reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
hana-developer-cli-tool-example mcp-server-development.instructions.md
Use when creating or updating MCP (Model Context Protocol) server components. Enforces TypeScript patterns, JSON-RPC communication rules, tool/resource/prompt registration, and integration with CLI command metadata. Ensures MCP server follows protocol requirements and maintains consistency with the CLI infrastructure.
phone-a-friend CLAUDE.md
Instructions for freibergergarcia/phone-a-friend, a project described as: CLI relay for Claude Code, Codex, and OpenCode. Delegate to Antigravity, Claude, Codex, Gemini, Ollama, or OpenCode with streaming, sessions, and multi-agent teams.
spellbook AGENTS.md
Instructions for majiayu000/spellbook, covering spellbook agent contract, routing, scope rules, threads long-run guardrails and validation.
open-supermarkets AGENTS.md
Instructions for abracadabra50/open-supermarkets, covering agent integration guide, supported frameworks, quick integration, 1. add as skill and 2. agent calls commands.
helm copilot-instructions.md
Instructions for PetePeter/helm, covering gamepad-cli-hub — copilot instructions, project purpose, system overview, data flow pipeline and key controls.
flyto-core CLAUDE.md
Instructions for flytohub/flyto-core, covering claude notes, cross-agent handoff and shared code intelligence.