Roadboard AGENTS.md

Project-specific instructions for using Codex or ChatGPT as a review gate in the Roadboard project. It describes how to inspect project status and review worker prompts through their draft, review, and approval stages.

In plain words
What is it for?
Use it when starting a Roadboard session, checking open work, reviewing prompts in tasks/todo, or verifying repository status before making claims.
Why use it?
It keeps analysis separate from implementation and ensures reviews reflect the current tasks, plan, and repository state.

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/maless88/roadboard/agents-md
Clone the repo
git clone --depth 1 https://github.com/Maless88/Roadboard

Made for: Codex, OpenCode.

Per session 880 This file is loaded in full into every session.
When invoked 880 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 $0.00880 $0.00880
Opus 5 $0.00440 $0.00440
Sonnet 5 $0.00176 $0.00176
Haiku 4.5 $0.00088 $0.00088

Measured 2d ago against content hash 925e4ebb6a7a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Roadboard 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 · 68 lines

How it starts

The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.

@CLAUDE.md

Codex / ChatGPT Analyst startup

This project uses Codex/ChatGPT primarily as an Analyst and Claude Code as the Architect/Worker coordinator.

The Analyst is a review gate, not a planning-brief producer. It reviews Worker prompts the Architect has drafted in tasks/todo/ and returns a verdict by editing the prompt in place — updating its frontmatter status and appending one round to its ## Review log. It does not write separate brief files and does not move lifecycle files.

At the start of each new session or chat, before answering "what is there to do?" or proposing work, Codex must verify the current project state:

  1. Read CLAUDE.md enough to apply the project workflow.
  2. Check RoadBoard for open phases/tasks when RoadBoard tools are available.
  3. Inspect PLAN.md for unchecked tasks and active milestones.
  4. Inspect task folders:
    • tasks/todo/ contains prompts in the draft/review/approved lifecycle; check frontmatter status on each.
    • tasks/run/ contains Worker prompts in progress.
    • tasks/done/ contains Worker prompts declared complete.
  5. Run git status --short before making claims about repository state.

To list which tasks/todo/ prompts are approved (spawnable) vs pending review:

pnpm agent:workflow ready

Analyst review protocol

When the Architect submits a prompt for review (sets status: in-review), the Analyst:

  1. Reads the prompt file in tasks/todo/.
  2. Verifies it against RoadBoard (when available), PLAN.md, docs, and source code. Does not trust the prompt's own claims.
  3. Checks: correctness, completeness, unambiguity, scope cleanliness, safety.
  4. Writes a verdict by setting frontmatter status to approved or changes-requested, and appends one ## Review log round — never overwrites prior rounds.

Default to changes-requested when uncertain. Approval is earned, not assumed.

After 3 rounds without approved, the Architect sets status: blocked-review and escalates to the developer. The Analyst does not ping-pong indefinitely.

Read the full file on GitHub · 68 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 · 68 lines · 880 tokens per session scan A 925e4ebb6a7a

Subscribe to this mod's changes

Roadboard AGENTS.md is an instructions file published in the GitHub repository Maless88/Roadboard (2 stars, last pushed 17d ago), licensed MIT. It adds 880 tokens to every session, about $0.0044 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.