AlexandrAI AGENTS.md
agents-md

Gumroad Product Planner Agent Guide

Project-specific AGENTS.md guidance for the Gumroad product planner role, schema-ready product.json, premium offer rules, page copy structure, and visual handoff.

AGENTS.md

Guidance for coding agents working on the Gumroad product planner role.

Project Overview

The product planner turns the CEO decision and market brief into products/<slug>/product.json. It owns commercial structure: title, subtitle, buyer, promise, price, file inventory, simple Gumroad HTML, refund, support, disclaimer, cover prompt, alt text, and page design notes.

Source Files

  • agents/product-planner.md defines the role.
  • skills/product-page-copywriter/SKILL.md defines the product.json and description contract.
  • skills/demo-driven-product-page/SKILL.md defines the active page pitch pattern.
  • schemas/product.schema.json defines required product fields.
  • roadmap/product-roadmap.md defines brand position and premium offer standard.
  • agents/product-page-designer.md receives the page and visual handoff.

Setup Commands

bashUV_LINK_MODE=copy UV_PROJECT_ENVIRONMENT="$HOME/.cache/gumroad-codex-agent-venv" uv run python scripts/product_artifacts.py render products/<slug>/product.json

Rendering later verifies that planned sections and templates can become buyer-facing artifacts.

Change Rules For Agents

  • Keep one buyer segment, one AI-assisted workflow, one purchase promise, and one proof path.
  • Use literal searchable titles and stable hyphenated slugs.
  • Default premium AI execution kits to USD 29 unless the owner explicitly asks for a smaller entry offer.
  • Keep description_html to simple h2, p, ul, li, and strong tags.
  • Mention PDF, template, support, and disclaimer in the description.
  • Include exact file, page, skill, harness, tracker, or example proof when promised.
  • Keep consultant-built provenance visible without inventing clients, results, testimonials, or sales data.
  • Put platform refund expectations in refund_policy; do not clutter description_html unless needed.

Product JSON Rules

Required fields include slug, title, subtitle, audience, promise, summary, price_cents, currency, taxonomy_id, taxonomy_path, tags, description_html, sections, templates, refund_policy, disclaimer, support, originality_notes, cover_prompt, cover_alt_text, and page_design_notes.

Verification

bashUV_LINK_MODE=copy UV_PROJECT_ENVIRONMENT="$HOME/.cache/gumroad-codex-agent-venv" uv run python scripts/product_artifacts.py render products/<slug>/product.json
UV_LINK_MODE=copy UV_PROJECT_ENVIRONMENT="$HOME/.cache/gumroad-codex-agent-venv" uv run python scripts/product_artifacts.py qa products/<slug>

Run QA after content is complete. Planner changes should make later QA easier, not rely on QA to discover the offer.

Security Notes

Do not include secrets, buyer data, private account ids, private endpoints, or absolute local paths in product.json or page copy. Keep automation JSON credential-safe when planned.

PR Checklist

  • Product belongs to a roadmap lane.
  • Description opens with a buyer pain or purchase moment.
  • Exact inventory proof is visible.
  • Claims match planned files.
  • Cover prompt and alt text are specific.