end to end app store screenshot creation using AI
Contributing
Thanks for contributing to app-store-screenshots.
This repository is intentionally small, but changes still affect real agent behavior. Most contributions here change either:
README.md: how humans discover and install the skillskills/app-store-screenshots/SKILL.md: how coding agents actually behave
What Makes a Good Contribution
- Fixes a real workflow problem for people generating App Store screenshots
- Improves the quality or reliability of the generated output
- Makes the skill easier to use across agents and environments
- Keeps the skill opinionated, but still user-driven
Scope Guidelines
Good fit:
- Better export reliability
- Better prompt flow and requirements gathering
- Stronger copy/design guidance
- Better contributor ergonomics
- Clearer installation or usage docs
Usually not a fit:
- Hardcoded brand styles
- Repo-specific assumptions that do not generalize
- Large framework additions outside the skill's purpose
- Changes that make the skill significantly more verbose without improving outcomes
Before Opening a PR
- Check open PRs to avoid overlapping work.
- Read both
README.mdandskills/app-store-screenshots/SKILL.md. - Keep user-facing docs and skill behavior aligned when applicable.
Testing Changes
The canonical product is skills/app-store-screenshots/template, which ships the editor and its verification suite. Start with:
cd skills/app-store-screenshots/template
bun install --frozen-lockfile
bun run typecheck
bun run build
bun run test:e2e
SCREENSHOTS_E2E_PRODUCTION=1 bun run test:e2e
The Tester Army configuration launches a disposable template copy and uses installed Google Chrome. Every scaffold contains the same configuration, tests, legacy regression harnesses and CI workflow. Read the flow matrix for exact coverage and residual gaps.
The original scripts/{bug-bash,ui-bug-bash,api-bug-bash}.cjs commands remain compatibility entry points. Only run them against a disposable editor server; they write its project and upload assets. The API and export harnesses must run sequentially. Set BUG_BASH_FILTER for one browser check.
For README-only changes
- Confirm installation instructions still make sense
- Confirm example paths and commands are copy-pasteable
For SKILL.md changes
Validate the skill against at least one realistic scenario:
- Start from an empty folder
- Ask an agent to generate App Store screenshots for a sample app
- Confirm the skill:
- asks the required discovery questions first
- preserves the "screenshots are ads, not docs" principle
- keeps the generator architecture coherent
- produces export instructions that are internally consistent
Strongly Recommended
Include in your PR description:
- the prompt you used to test the skill
- what behavior changed
- what stayed intentionally unchanged
Authoring Guidelines
- Prefer compact, high-signal instructions over long prose
- Keep examples realistic and production-oriented
- Avoid duplicating large blocks between
README.mdandSKILL.mdunless the duplication helps different audiences - If a change modifies workflow expectations, document it in both files when relevant
Pull Request Checklist
Before submitting, verify:
- the change solves a concrete problem
- the wording is clear for both humans and agents
- README and SKILL instructions do not contradict each other
- the PR description explains why the change is useful