The Agent Skills Directory

[PROMPT_INJECTION]: The skill performs an automated analysis of project source code, data models, and configuration files to infer Kafka schemas and categorize applications. It lacks explicit boundary markers or instructions to disregard natural language commands embedded within the data it ingest (e.g., inside code comments, string literals, or README files).
Ingestion points: The agent reads arbitrary files from the target repository using Glob, Grep, and Read operations during Phases 1, 2, and 3.
Boundary markers: There are no specific delimiters or "ignore embedded instructions" warnings configured in the workflow when processing untrusted content.
Capability inventory: The skill has Write permissions to create new files, including reports, schemas, and automation scripts.
Sanitization: No explicit sanitization or filtering of external content is mentioned before the data is interpolated into the generated reports or configuration files.
[COMMAND_EXECUTION]: The skill is designed to generate executable shell scripts (e.g., terraform/import.sh in Phase 6) and Terraform configuration files (.tf). While the agent does not execute these files itself, it generates and places executable code into the user's workspace based on data parsed from the repository. This represents a potential risk if the inferred data is manipulated to inject malicious commands into the generated scripts.

kafka-schema-registry