Reusable product-review and product-rules skills for AI coding assistants, including Codex.
This repository helps teams turn vague product feedback into repeatable methods that an assistant can reuse across projects.
Use this repository when you need to:
- review related screens for structural, semantic, interaction, or visual drift
- validate whether a fix survives the full user flow
- turn repeated UX decisions into durable product rules
The goal is simple: spend less time re-explaining your review method and get more consistent product reasoning from your assistant.
Most users only need this flow:
- Read this
README.md. - Pick one skill for the current task.
- Either install that skill into Codex or ask your assistant to follow its
SKILL.md.
Repository URL:
https://github.com/rushairer/ai-product-skills
If your assistant supports repo-aware workflows, pointing it at this repository and asking it to read README.md first is usually enough.
Use for generic sibling-screen review.
Best for:
- comparing related pages
- finding shared page skeleton drift
- deciding whether a difference is intentional or accidental
Use for domain-aware review in mature product families.
Best for:
- products with stable in-domain terminology
- repeated page families with specialized meaning
- cases where generic wording would flatten important distinctions
Use for end-to-end path validation.
Best for:
- checking whether a user choice keeps the same meaning through navigation, runtime, persistence, recovery, history, and re-entry
- verifying fixes after UI consistency or state-model work
Use for extracting and maintaining durable product rules.
Best for:
- creating or updating
APP_RULES.md,FEATURE_RULES.md, or similar files - separating reusable principles from app-specific behavior
In most projects:
- Start with
product-ui-consistency-review-core. - Switch to
product-ui-consistency-review-specializedonly if product-specific terminology materially affects the review. - Use
product-flow-validationafter consistency fixes when the remaining question is whether meaning survives the full path. - Use
app-rules-architectwhen the same review outcomes keep recurring and should become rules.
Install the beginner set:
./scripts/install_for_codex.shThat installs:
product-ui-consistency-review-coreproduct-flow-validation
Install every skill:
./scripts/install_for_codex.sh --allInstall one or more specific skills:
./scripts/install_for_codex.sh product-ui-consistency-review-core app-rules-architectAfter installation, restart Codex.
If you prefer to copy folders yourself:
mkdir -p ~/.codex/skills
cp -R skills/product-ui-consistency-review-core ~/.codex/skills/
cp -R skills/product-flow-validation ~/.codex/skills/For the full set:
cp -R skills/product-ui-consistency-review-specialized ~/.codex/skills/
cp -R skills/app-rules-architect ~/.codex/skills/If your assistant cannot install Codex skills natively, use this repository as a structured prompting toolkit:
- Point the assistant to this repository.
- Ask it to read
README.md. - Point it to exactly one relevant
SKILL.md. - Tell it to use that skill as the working method for the task.
If you want ready-made prompts:
./scripts/print_bootstrap_prompts.shFor more concrete examples, see docs/USAGE_EXAMPLES.md.
skills/
app-rules-architect/
product-flow-validation/
product-ui-consistency-review-core/
product-ui-consistency-review-specialized/
templates/
APP_RULES_TEMPLATE.md
docs/
RULES_AND_SKILLS_README.md
SKILL_BOUNDARIES.md
USAGE_EXAMPLES.md
Use these documents by purpose:
- README.md: onboarding and skill routing
- docs/SKILL_BOUNDARIES.md: when two skills seem similar
- docs/RULES_AND_SKILLS_README.md: how skills, templates, and rules fit together
- docs/USAGE_EXAMPLES.md: concrete prompting examples
- templates/APP_RULES_TEMPLATE.md: reusable starting point for app rules
- CONTRIBUTING.md: contribution expectations
scripts/install_for_codex.sh: install the default, full, or custom skill set into~/.codex/skillsscripts/print_bootstrap_prompts.sh: print ready-to-copy prompts for common assistant setups
See CONTRIBUTING.md.
MIT