The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains hardcoded credentials used as placeholders in local development examples.
Evidence: The SKILL.md file uses root for both username and password in the server start commands and the db.signin authentication call.
[INDIRECT_PROMPT_INJECTION]: The provided code templates for database interaction use raw string interpolation for queries, creating a potential surface for injection attacks if an agent populates these strings with unvalidated user input.
Ingestion points: Agent-constructed SurrealQL queries in SKILL.md and references/embedded.md that may interpolate external data.
Boundary markers: Absent in the provided Python code examples.
Capability inventory: Full CRUD (Create, Read, Update, Delete) access to the database instance.
Sanitization: The examples do not demonstrate sanitization or the use of parameterized queries.

surrealdb-python