agent-engineering-standards AGENTS.md

A set of public conventions for coding agents working in repositories. It defines ways to write requirements, protect personal or secret data, and describe the inputs, outputs, and stopping conditions of multi-step agent tasks.

In plain words
What is it for?
Use it to write acceptance criteria in EARS form, route sensitive data through private workflows, and define clear contracts for agent steps.
Why use it?
It makes agent-assisted software work easier to review and audit, while reducing the chance that private data or unclear requirements are handled incorrectly.

Instructions file for CodexOpenCode

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 instructions/onesimplecode/agent-engineering-standards/agents-md
Clone the repo
git clone --depth 1 https://github.com/onesimplecode/agent-engineering-standards

Made for: Codex, OpenCode.

Per session 7,592 This file is loaded in full into every session.
When invoked 7,592 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original 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 $0.07592 $0.07592
Opus 5 $0.03796 $0.03796
Sonnet 5 $0.01518 $0.01518
Haiku 4.5 $0.00759 $0.00759

Measured yesterday against content hash f3c416b4833a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-engineering-standards 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 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.

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.

AGENTS.md · 639 lines

How it starts

The opening of the file, as written. The whole thing — 639 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md — Public Agent Conventions

These conventions are the portable, tool-neutral layer of the AI Engineering Standards. They can be read by Claude Code, Cursor, Codex, Gemini CLI, or any agent that accepts repository instructions.

The goal is not to make agents more verbose. The goal is to make agent-assisted software work bounded, reviewable, and auditable.

Core Rules

Requirements Syntax

Write requirements in EARS style when they become acceptance criteria:

When <trigger>, the system shall <observable behavior>.

Avoid vague words like "should" once a requirement is active.

Data Routing (TR-SEC-003)

Data that can identify a person, expose a secret, or reveal private business context must be handled by a local/private workflow unless an ADR documents an explicit exception.

Cloud-backed agents may work on public docs, public code, synthetic examples, and low-risk drafting. They must not receive secrets or private datasets.

Loop Contracts (TR-AGT-003)

Every multi-step agent node declares four fields before implementation:

  1. Input schema — expected state or data.
  2. Output schema — produced state or data.
  3. Exit condition — observable evidence that the node is done.
  4. Resource budget — max iterations, token budget, or wall-clock timeout.

Missing any field means the design is incomplete.

The exit condition must be verified by deterministic evidence when the agent changes persistent state, writes files, sends messages, or calls tools with side effects (TR-TEST-006).

See examples/engine-interface/ for a concrete reference implementation: a SearXNG-inspired multi-source polling pattern where source_name (identity), default_timeout (budget), a never-raising fetch() (exit condition), and a normalized list[Result] (output schema) map directly onto the four fields above.

MCP Tool Annotations (TR-AGT-003, field 5)

When a node is exposed as an MCP tool, declare four hint flags describing its blast radius. These are advisory hints to MCP clients (Claude Code, Cursor, opencode) — the MCP protocol does not enforce them, so declare them accurately regardless.

Read the full file on GitHub · 639 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. yesterday First seen · 639 lines · 7,592 tokens per session scan A f3c416b4833a

Subscribe to this mod's changes

agent-engineering-standards AGENTS.md is an instructions file published in the GitHub repository onesimplecode/agent-engineering-standards (3 stars, last pushed 4d ago), licensed MIT. It adds 7,592 tokens to every session, about $0.0380 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.