The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied application metadata (such as names, descriptions, and developer info) and incorporates it into generated manifests, XML MetaInfo files, and shell scripts.
Ingestion points: Information gathered in 'Step 1' (names, summaries, descriptions, URLs) is used to populate templates across several files.
Boundary markers: The skill does not explicitly define delimiters for untrusted user input within the prompts.
Capability inventory: The skill is designed to create multiple configuration files and executable shell scripts (.sh) for application sandboxing.
Sanitization: While Step 2 provides strict validation rules for Application IDs, it does not provide specific instructions for sanitizing or escaping free-text fields in XML or Shell contexts.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for obtaining official development tools and shared components from reputable sources.
Fetches the official flatpak-builder-tools from the project's GitHub organization.
References official shared-modules from the Flathub GitHub organization.
Recommends the use of community-standard tools like flatpak-go-mod for specific language support.
[DYNAMIC_EXECUTION]: The skill automates the generation of shell launcher scripts and installation scripts (apply_extra) based on provided templates. This is a standard and necessary component of the Flatpak packaging process.

flatpak-flathub