Grace AGENTS.md

A project instruction file for the ScottArbeit/Grace codebase. It documents setup requirements, useful commands, issue tracking, development rules, and Markdown guidelines.

In plain words
What is it for?
Use it when working on this repository to bootstrap the environment, run focused or broader validation, follow its development process, and find more specific instructions.
Why use it?
It gives coding agents and contributors the project-specific rules needed to make and validate changes consistently.

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/scottarbeit/grace/agents-md
Clone the repo
git clone --depth 1 https://github.com/ScottArbeit/Grace

Made for: Codex, OpenCode.

Per session 3,990 This file is loaded in full into every session.
When invoked 3,990 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 $0.03990 $0.03990
Opus 5 $0.01995 $0.01995
Sonnet 5 $0.00798 $0.00798
Haiku 4.5 $0.00399 $0.00399

Measured 3d ago against content hash bf029e8b2553, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Grace 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 3d 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 · 122 lines

How it starts

The opening of the file, as written. The whole thing — 122 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Instructions

Other AGENTS.md files exist in subdirectories, refer to them for more specific context.

Agent Quickstart (Local)

Prerequisites:

  • PowerShell 7.x
  • .NET 10 SDK
  • Docker Desktop (required for -Full)

Commands:

  • pwsh ./scripts/bootstrap.ps1
  • Run the smallest meaningful focused proof for the changed slice.

Use pwsh ./scripts/validate.ps1 -Fast as an optional broad local preflight and -Full for local integration reproduction or diagnosis. GitHub Validate is the required broad gate for the current pull-request revision. Optional: pwsh ./scripts/install-githooks.ps1 adds a lightweight pre-commit staged-diff check; focused tests and formatting remain explicit responsibilities.

More context:

  • src/AGENTS.md
  • skills/grace/SKILL.md for Agent Skills-compatible clients; use it as the portable Grace context router after reading repo-local instructions.
  • src/docs/ASPIRE_SETUP.md
  • src/docs/ENVIRONMENT.md

Issue Tracking

  • Use GitHub issues and pull requests as the active coordination surface for implementation work.
  • For non-trivial work, follow docs/Development process.md: create or confirm a GitHub issue, declare owned paths, create an issue-owned branch/worktree, validate in focused slices, commit after each completed slice, and record docsimpact and skipped validation.
  • For multi-step implementation plans, create an epic parent issue, but initially create only the earliest tracer and prerequisites proven necessary by current evidence. Use native GitHub parent relationships and a compact dependency map for created work. Do not pre-create a broad horizontal issue forest. Re-plan and create later child issues after the tracer runs. Use the concrete addSubIssue GraphQL workflow in docs/Development process.md for each created child.
  • When planning a feature or epic, and when creating issue or pull request bodies, include why the change matters for Grace and its users. Use that purpose to help implementation agents make better local decisions when the plan leaves a gap or an acceptance criterion is ambiguous.
  • When creating the epic and child issues from PowerShell, avoid one giant inline script that embeds every issue body.
  • For large issue batches, go directly to a short-lived generator script checked into the worktree or written in a temp directory, run that script to emit one temporary Markdown body per issue, lint those files, then call gh issue create --body-file <path> for each issue. Do not paste large scripts through an interactive shell or pass them as a single pwsh -Command string; that wastes time, floods the transcript, and can hit Windows command-length limits. After issue creation, patch the epic body with the real child issue numbers and create the native relationships with GraphQL addSubIssue.
  • Keep epics as durable product and planning records, not long-running agent conversations. Use fresh epic-checkpoint sessions before the first child, after each merged child, and at epic closure. Select each child's delivery mode explicitly. Default to a mainline slice from current origin/main when the child is independently correct and safe or inert until consumed. Use an epic integration branch only when the owner approves a real composition need, CI coverage, refresh policy, delivery-delta review, and final epic-to-main release candidate. Do not create an epic branch merely because an issue has an epic parent.
  • Use one checkpoint agent by default; at most two read-only scouts may answer concrete independent questions with explicit fork_turns = "none". Checkpoints never spawn implementation workers.
  • Every pull request must link its related GitHub issue in the PR body. When a PR targets the default branch and should close an issue, use one of GitHub's supported closing keywords: close, closes, closed, fix, fixes, fixed, resolve, resolves, or resolved. Use docs/Development process.md for default-branch versus epic-branch wording so links stay traceable without relying on epic-branch auto-close behavior.

Read the full file on GitHub · 122 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. 3d ago First seen · 122 lines · 3,990 tokens per session scan A bf029e8b2553

Subscribe to this mod's changes

Grace AGENTS.md is an instructions file published in the GitHub repository ScottArbeit/Grace (587 stars, last pushed 3d ago), licensed MIT. It adds 3,990 tokens to every session, about $0.0199 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.