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/kernlang/agon/agents-mdgit clone --depth 1 https://github.com/KERNlang/agonWhat 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 | $0.03141 | $0.03141 |
| Opus 5 | $0.01571 | $0.01571 |
| Sonnet 5 | $0.00628 | $0.00628 |
| Haiku 4.5 | $0.00314 | $0.00314 |
Grade A, and why
agon 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 2d 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 — 239 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Agon AI
Any AI can join. They compete. You ship.
Agon is a multi-AI orchestration framework. Registered engines compete via forge, debate via tribunal, and ideate via brainstorm. Cesar orchestrates from live capability, availability, confidence, and task evidence. Canonical cross-agent doctrine: ~/.claude/AGENTS.md; Agon-specific global conventions: ~/.agon/AGON.md.
Using Agon From Codex
Codex agents should use Agon when the task benefits from multiple engines, adversarial review, team competition, or confidence-weighted ideation, but interactive chat must never auto-start these modes just because the user mentions "brainstorm", "tribunal", "forge", or "review". In the Agon UI, orchestration starts only from explicit slash commands (/brainstorm, /tribunal, /forge, /review, etc.) or from explicit CLI/MCP calls. If the user is speaking normally, suggest the slash command instead of starting the workflow.
For non-interactive Codex shell work, the fastest path is the shell bridge:
agon call brainstorm "What approaches should we consider for this migration?"
agon call tribunal "Should we ship this architecture?" --tribunalMode red-team --rounds 2
agon call forge "Implement the cache layer" --test "npm test"
agon call synthesis "Evolve this design doc into a concrete implementation plan" --swaps 2 --timeout 90
agon call review
If agon is not linked in the shell, build/link it first:
npm run build
npm run install:cli
Machine-readable callers should add --jsonl:
agon call brainstorm "Compare options for this refactor" --jsonl
Use --cwd <path> when the target repository is not the current working directory. Use --engines <comma-separated-registered-ids> only when the user explicitly requests manual pinning. Use --timeout <seconds> for long-running tasks.
External engines should always call Agon through the shell bridge, not direct model CLIs:
agon call <workflow> "<input>" [flags]
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.
- 2d ago First seen · 239 lines · 3,141 tokens per session scan A 5e9138449885
agon AGENTS.md is an instructions file published in the GitHub repository KERNlang/agon (11 stars, last pushed 7d ago), licensed MIT. It adds 3,141 tokens to every session, about $0.0157 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 2026-08-30.
Other instructions, from other repositories
multi-model-team CLAUDE.md
Claude Code instructions for WatchAndyTW/multi-model-team, covering multi-model-team — project guide for claude, ⚠️ backend invocation quirks, agy (gemini) — needs a tty → runs under node-pty, opencode — no tty, prompt on stdin, and it ignores cwd and codex — no tty needed.
neo-skills AGENTS.md
Instructions for Benknightdark/neo-skills, covering neo-skills agents.md, 1. project overview, 2. commands, 3. workflow and 4. forbidden antipattern redlines.
agy-bridge CLAUDE.md
Instructions for sshahzaiib/agy-bridge: You have agy-bridge MCP tools that delegate heavy work to the Antigravity CLI (Gemini). Delegation keeps large content OUT of your context — only answers come back. Prefer delegating over doing it yourself when.
agy-plugin-cc CLAUDE.md
Instructions for Vit129/agy-plugin-cc, covering agy-plugin-cc — claude code guide, agent memory, global-first rule, project context (auto-loaded every session) and project summary.
ag-sdd AGENTS.md
AGENTS.md instructions for stormlive-ai/ag-sdd, covering agents.md — anti-gravity spec-driven development (ag-sdd), 1. purpose & philosophy, 2. sdd workflow lifecycle, phase 1 — discovery (mandatory clarification gate) and phase 2 — spec generation.
apc.md AGENTS.md
Instructions for tk-wxy/apc.md, covering agent entry, route, task end and core.