The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents a framework that processes data from external, potentially untrusted sources (e.g., via HTTP steps or workflow arguments) and uses that data to perform sensitive operations such as database transactions and network requests. This creates a surface for indirect prompt injection if the ingested data contains instructions meant to influence the AI agent's logic.
Ingestion points: SKILL.md (workflow method arguments), references/step-basics.md (external API calls via httpGet).
Boundary markers: The provided code examples do not explicitly demonstrate the use of delimiters or instructions to ignore embedded natural language commands within processed data.
Capability inventory: The SDK provides broad capabilities including database access (JdbcStepFactory), network operations (HttpClient), and file system access.
Sanitization: The documentation does not specifically address sanitization techniques for preventing prompt injection when handling external data within workflows.
[EXTERNAL_DOWNLOADS]: The skill references official DBOS libraries (e.g., dev.dbos:transact) and utilizes official documentation and repository links (docs.dbos.dev, github.com/dbos-inc/*). These are verified vendor resources and are documented neutrally for application configuration purposes.

dbos-java