coding-agent-toolkit AGENTS.md

Project instructions for coding-agent workflows that guide work from an initial idea to a merged GitHub pull request. They also support non-code deliverables such as research reports or course material.

In plain words
What is it for?
Use them to align on a task, create a plan, split work into issues, pick the next issue, hand work off, continue later, and reflect progress in GitHub.
Why use it?
They provide a shared process for defining done, planning work, recording references and constraints, tracking progress, and continuing across sessions or coding tools.

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/stefan-jansen/coding-agent-toolkit/agents-md
Clone the repo
git clone --depth 1 https://github.com/stefan-jansen/coding-agent-toolkit

Made for: Codex, OpenCode.

Per session 1,341 This file is loaded in full into every session.
When invoked 1,341 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 $0.01341 $0.01341
Opus 5 $0.00671 $0.00671
Sonnet 5 $0.00268 $0.00268
Haiku 4.5 $0.00134 $0.00134

Measured 2d ago against content hash 67d9b1a5e98d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

coding-agent-toolkit 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.

AGENTS.md · 117 lines

How it starts

The opening of the file, as written. The whole thing — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.

coding-agent-toolkit

Contributor / maintainer notes. For the project introduction, see README.md.

Purpose

Provide a small set of host-neutral workflow steps — align, plan, plan-issues, next-issue, ship, handoff, continue — that take a piece of work from rough idea to done, identically on Claude Code and OpenAI Codex. The chain owns the structural work around producing the thing: clarifying what "done" means, breaking it into chunks, accumulating references and constraints as they surface, tracking progress, carrying memory across sessions and host swaps, and projecting state onto GitHub for review and version control. Local state for each piece of work lives in .workspace/work/<unit>/ (spec, plan, follow-up notes, references); GitHub is the public projection (milestone / issues / branches / PRs / Closes #N). The dominant deliverable is a merged PR; the chain drives non-code work too (a research report, a course module, a long-form post) since the projection doesn't care what the deliverable is. The actor is the agent; this repository supplies the step definitions and the file-shaped contract they share.

Cross-host primitive

The contract that makes the toolkit host-neutral is shared durable state under .workspace/, read natively by both Claude Code and Codex. There is no orchestrator. Either host runs a step; the step writes files; the other host picks the state up by reading the same files.

Path Role
AGENTS.md (this file) Canonical project instructions. Codex reads natively; Claude includes via CLAUDE.md.
CLAUDE.md One line — @AGENTS.md.
.workspace/memory/ Persistent project memory; load on demand.
.workspace/transitions/YYYY-MM-DD/HHMMSS.md /handoff output; /continue resumes from any of these on either host.
.workspace/work/ Active work units (specs, plans, references, follow-up notes).

Keep this contract tight. Never seed .claude/transitions/ or .codex/state/ for new work; never invent a run-as-other-host shortcut.

Read the full file on GitHub · 117 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. 2d ago First seen · 117 lines · 1,341 tokens per session scan A 67d9b1a5e98d

Subscribe to this mod's changes

coding-agent-toolkit AGENTS.md is an instructions file published in the GitHub repository stefan-jansen/coding-agent-toolkit (22 stars, last pushed 16d ago), licensed MIT. It adds 1,341 tokens to every session, about $0.0067 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.

Related

Other instructions, from other repositories