The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from external sources (e.g., GitHub, Snowflake) into Dagster assets. This introduces a theoretical risk of indirect prompt injection if the agent is later asked to analyze or act upon the content of the ingested data.
Ingestion points: External dlt sources as described in SKILL.md and references/component.md.
Boundary markers: Not explicitly defined in the skill instructions.
Capability inventory: Includes network operations for data ingestion and local file system interactions for pipeline state, as detailed in the orchestration workflows.
Sanitization: Relies on the default security posture of the dlt and Dagster libraries.
[EXTERNAL_DOWNLOADS]: The skill suggests the installation of additional skills from external repositories to extend its capabilities.
It references the dagster-integrations skill from the official GitHub organization of Dagster (dagster-io), which is a well-known technology provider.
It references the dlt-skill from the author's own repository (untitled-data-company/data-skills).

dlt-dagster