agentfield AGENTS.md

Workspace instructions that record user preferences and project facts for AgentField, an open-source system for running software agents. They guide decisions about APIs, the user interface, observability, configuration, and desktop features.

In plain words
What is it for?
Use them when designing or reviewing AgentField features, especially desktop agent management, activity views, observability, configuration files, and open-source versus enterprise responsibilities.
Why use it?
They keep coding-agent work aligned with established product boundaries and design choices. This reduces repeated decisions and prevents changes that conflict with the project’s direction.

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/agent-field/agentfield/agents-md
Clone the repo
git clone --depth 1 https://github.com/Agent-Field/agentfield

Made for: Codex, OpenCode.

Per session 471 This file is loaded in full into every session.
When invoked 471 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 $0.00471 $0.00471
Opus 5 $0.00235 $0.00235
Sonnet 5 $0.00094 $0.00094
Haiku 4.5 $0.00047 $0.00047

Measured 2d ago against content hash cfaf86834c50, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agentfield 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 · 18 lines

How it starts

The opening of the file, as written. The whole thing — 18 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Learned User Preferences

  • Open-source AgentField should prioritize stable APIs and primitives so integrators can build advanced observability themselves; large packaged business or fleet observability belongs in Enterprise.
  • The embedded OSS UI should stay a lightweight convenience layer, not the primary surface for org-wide analytics or governance-heavy views.
  • Developer-facing observability belongs in OSS; deeper reliability and governance programs may span OSS and Enterprise.
  • Avoid empty or placeholder PRs when stacking branches; prefer draft PRs with real implementation, then thorough review before marking ready.
  • When designing or documenting control plane behavior, treat YAML configuration (config/agentfield.yaml and AGENTFIELD_CONFIG_FILE) as a first-class surface alongside environment variables.
  • AgentField Desktop targets GitHub-comfortable developers (not infra experts); primary jobs are installing agent nodes from GitHub and seeing runs/cost as a local sub-harness for coding agents.
  • Desktop UI should use shared theme tokens rather than hardcoded page styles; treat Agents as a marketplace-style library (installed agents + add), and design Activity for high-volume dense/filterable runs rather than large cards.
  • Locked desktop decisions: gold/amber accent; cold-launch to Home when agents exist (add/empty flow when none); usage totals on Home plus Activity per-row when the API allows; keep the update banner across views.

Learned Workspace Facts

  • Monorepo: Go control plane in control-plane/, SDKs in sdk/, embedded admin UI in control-plane/web/client/, Electron desktop app in desktop/.
  • Agent-node manifests (agentfield-package.yaml) carry a config_version (schema version, e.g. v1; absent = v0) that is separate from the node's own version:. Bump config_version only for breaking format changes, never for additive fields. The single reader is packages.ParsePackageMetadata (control-plane/internal/packages/installer.go); the authoring contract lives in docs/installing-agent-nodes.md.
  • Desktop design/product specs live in DESIGN.md and PRODUCT.md.
  • Desktop featured-catalog copy is maintained in desktop/src/shared/catalog.ts; post-install descriptions come from each agent's agentfield-package.yaml (marketplace cards do not fetch YAML live).

Read the full file on GitHub · 18 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 · 18 lines · 471 tokens per session scan A cfaf86834c50

Subscribe to this mod's changes

agentfield AGENTS.md is an instructions file published in the GitHub repository Agent-Field/agentfield (2,539 stars, last pushed 2d ago), licensed Apache-2.0. It adds 471 tokens to every session, about $0.0024 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.