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/thuandiep1707/codex-harness-javascript/agents-mdgit clone --depth 1 https://github.com/thuandiep1707/codex-harness-javascriptWhat 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 | $0.04411 | $0.04411 |
| Opus 5 | $0.02205 | $0.02205 |
| Sonnet 5 | $0.00882 | $0.00882 |
| Haiku 4.5 | $0.00441 | $0.00441 |
Grade A, and why
codex-harness-javascript 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 — 378 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Codex Multi-Agent Delivery System
This repository is the control project for a zero-setup Codex multi-agent workflow. The product repository is a separate project opened in the same Codex workspace. Keep this repository selected as primary so this file is loaded at the start of every chat.
System truths
Use exactly these durable sources of truth:
- Control repository: workflow entry points, agent behavior, internal capabilities, rules, protocols, and Codex configuration.
- Working project
.docs/: human-owned product truth. - Jira: work state and durable execution context.
- Working project source: implementation truth.
Chat history is never workflow truth. Do not create .plans/, .progresses/, .agent/, or another runtime workflow database in the product repository.
Public workflow registry vs internal capabilities
Keep user-facing workflow discovery intentionally small.
Public workflows
Only packages under .agents/skills/ are user-facing $ entry points discoverable by Codex.
Current frontend workflows:
$frontend-delivery: run frontend work continuously from the smallest valid entry through analysis, Jira orchestration, specialist execution, testing, runtime cleanup, child-agent cleanup, and final acceptance.$frontend-planning: analyze and create/reconcile the Jira work graph, then stop before specialist execution.
Do not place agent implementation knowledge under .agents/skills/.
Internal capabilities
Reusable agent knowledge lives under .agents/capabilities/ and is private to the owning agent. Internal capabilities must not be exposed as user-facing $ commands and must not be loaded globally.
Examples:
.agents/capabilities/common/discover-project-stack/
.agents/capabilities/frontend/plan-frontend-work/
.agents/capabilities/frontend/shadcn/
.agents/capabilities/frontend/nextjs-tanstack-query/
.agents/capabilities/frontend/testing/
An agent may load an internal capability only when:
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 · 378 lines · 4,411 tokens per session scan A 62e9d03f0d2e
codex-harness-javascript AGENTS.md is an instructions file published in the GitHub repository thuandiep1707/codex-harness-javascript (7 stars, last pushed 4d ago), licensed MIT. It adds 4,411 tokens to every session, about $0.0221 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
Agent-Harness-Kit AGENTS.md
Instructions for Eduardo-Salvador/Agent-Harness-Kit, covering agent harness kit — operational map, request-routing gate — before all harness ceremony, direct-trivial fast path, vibe fast path and context pressure and graph-only tasks.
Agent-Harness-Kit CLAUDE.md
Instructions for Eduardo-Salvador/Agent-Harness-Kit, a project described as: An installable harness for Codex and Claude Code that gives coding agents durable context, scoped task graphs, bounded execution, and a clear path from idea to verified delivery.
crewplane AGENTS.md
Instructions for crewplaneai/crewplane, covering agents.md, repo facts, read first, repo map and project heuristics.
awesome-harness-engineering CLAUDE.md
Claude Code instructions for ai-boost/awesome-harness-engineering, a project described as: Awesome list for AI agent harness engineering: tools, patterns, evals, memory, MCP, permissions, observability, and orchestration.
crewplane CLAUDE.md
Instructions for crewplaneai/crewplane, a project described as: Run reviewable, resumable coding-agent workflows across Claude Code, Codex, Gemini, GitHub Copilot CLI, or any CLI. You define the stages in Markdown; Crewplane runs the workflow, agents do the work, and every handoff stays on disk.
crewplane GEMINI.md
Instructions for crewplaneai/crewplane, a project described as: Run reviewable, resumable coding-agent workflows across Claude Code, Codex, Gemini, GitHub Copilot CLI, or any CLI. You define the stages in Markdown; Crewplane runs the workflow, agents do the work, and every handoff stays on disk.