The Agent Skills Directory

[SAFE]: The skill is a set of instructional guidelines for developers to create secure and uniform API endpoints. It defines a "Business Action" contract that mitigates common security risks.
[SAFE]: Rule R6 explicitly addresses privilege management and authorization by mandating the use of SECURITY DEFINER with a restricted search_path to prevent search path hijacking. It further requires server-side identity resolution to prevent attackers or LLMs from forging user identities via input arguments.
[SAFE]: The template provided for new actions includes validation steps (Rule R4) and idempotency checks (Rule R5), which are standard defenses against race conditions and invalid data state changes.
[INDIRECT_PROMPT_INJECTION]: The skill defines how to ingest data from untrusted sources (p_body). While this creates a potential surface for indirect injection, the skill's primary purpose is to provide the defensive boundaries and validation rules necessary to handle such input safely.
Ingestion points: p_body parameter in the api.crear_<noun> function (SKILL.md).
Boundary markers: Rule R6 provides a logical boundary by preventing the use of caller-supplied arguments for identity.
Capability inventory: SQL INSERT and SELECT operations within defined domain schemas.
Sanitization: Validation logic and server-side identity resolution are required before processing input.

rpc-api-contract