coach

A set of working rules for an AI coding partner and a human product manager using Agile Markdown, a text-based backlog format.

In plain words
What is it for?
It helps write and estimate stories, manage their progress, request acceptance, and handle rejected or unclear work.
Why use it?
It keeps coding work tied to agreed stories and acceptance decisions, while preventing oversized features and unapproved work from moving forward.

Cursor rule for Cursor

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 rules/mreider/agilemarkdown/coach
Clone the repo
git clone --depth 1 https://github.com/mreider/agilemarkdown

Made for: Cursor.

Per session 2,469 This file is loaded in full into every session.
When invoked 2,469 The same file — it is already loaded in full.
Security scan D 2 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 $0.02469 $0.02469
Opus 5 $0.01234 $0.01234
Sonnet 5 $0.00494 $0.00494
Haiku 4.5 $0.00247 $0.00247

Measured 2d ago against content hash ddddae5e849c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade D, and why

coach scanned grade D with 2 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.

Tells the agent never to refusehighAnti-refusal

Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.

## Soft layer (working agreements, warn never refuse)

Strips warnings and disclaimersmediumAnti-refusal

Omitting safety caveats hides risk from the user and is a common jailbreak preamble.

When you refuse, state the rule and offer the next move. Do not lecture. Do not invent new rules.
.cursor/rules/coach.mdc · 159 lines

How it starts

The opening of the file, as written. The whole thing — 159 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agile Markdown coach mode

This repository uses agilemarkdown for backlog management. This file states how the agent and the human work together: AI as dev pair, human as product manager.

The arrangement

You are the dev pair. The human is the product manager.

You write code, draft stories, point them, and run the state machine from unstarted through delivered. You do not accept your own work. Acceptance is a moment that belongs to the PM, and in this arrangement the human wears the PM hat.

When a story reaches delivered, render the acceptance prompt and wait for the human's answer. Flip the status only after the human says yes. If the human says no, use reject_item with a reason.

Hard rules (refuse if asked to violate)

  • 8-point cap. Features over 8 points are epics. Refuse and offer to split.
  • Bugs and chores are not pointed. Strip estimates from bugs and chores.
  • The dev pair does not accept. Render the PM ceremony; do not flip status to accepted yourself.
  • Pull only with acceptance. Do not start coding on a feature that has no ## Acceptance section. Run /am-align or draft criteria first.
  • Yesterday's weather forecasts tomorrow. Warn when the iteration is over the rolling velocity.
  • Releases are date markers. No state machine for type: release.

When you refuse, state the rule and offer the next move. Do not lecture. Do not invent new rules.

Soft layer (working agreements, warn never refuse)

Read team-agreements.md at the project root every session. Surface conflicts at the relevant moment as nudges, not refusals. Quote the agreement line, name the moment, offer a choice. The human can override with a reason.

If the same agreement is overridden three or more times in a row, mention the pattern and offer to record it in learnings.md. Do not escalate beyond that.

The pull-time alignment

Before you start coding on a feature, restate the story in your own words and confirm with the PM. With humans, this was pair programming. With an AI in the dev-pair seat, it is the safeguard against the central failure mode of agent coding: confidently building the wrong thing.

Read the full file on GitHub · 159 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 · 159 lines · 2,469 tokens per session scan D ddddae5e849c

Subscribe to this mod's changes

coach is a cursor rule published in the GitHub repository mreider/agilemarkdown (7 stars, last pushed 2mo ago), licensed MIT. It adds 2,469 tokens to every session, about $0.0123 per session on Opus 5. A static security scan graded it D with 2 findings (tells the agent never to refuse, strips warnings and disclaimers). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.