The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from TYPO3 records and rendered HTML to generate and audit structured data, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads data from TYPO3 records (such as EXT:news detail records, Content Blocks, and domain records), site configuration, and existing application/ld+json producers as described in the SKILL.md and typo3-integration.md files.
Boundary markers: The instructions emphasize maintaining strict parity between visible content and structured values, which serves as a logical boundary for expected data content.
Capability inventory: The agent has access to DDEV, Composer, PHP, and the TYPO3 CLI to perform inventories, audits, and installations.
Sanitization: The skill explicitly mandates the use of a serialization boundary, safe JSON encoding for all values, and escaping editor-controlled text to prevent XSS or script injection in the final output, as detailed in the 'Implement in TYPO3' and 'Rendering rules' sections.
[COMMAND_EXECUTION]: The skill instructs the agent to use DDEV for executing Composer, PHP, and TYPO3 commands.
These operations are used for legitimate project inventory, package management, and verification tasks within the TYPO3 development environment.
[EXTERNAL_DOWNLOADS]: The skill suggests installing the brotkrueml/schema extension from its official repository and Packagist.
Fetches the extension from the official GitHub repository (brotkrueml/schema) and Packagist registry using Composer.
The instructions emphasize verifying package versions and constraints against the local installation before proceeding with the installation.

typo3-structured-data