The Agent Skills Directory

[SAFE]: No security issues detected. The skill focuses on providing architectural advice and generating configuration templates (SQL, IaC) for multi-tenant SaaS environments.
[PROMPT_INJECTION]: No prompt injection or behavior override patterns were found. The instructions are focused on the stated purpose of multi-tenant design.
[DATA_EXFILTRATION]: No evidence of data exfiltration or unauthorized network operations. The skill does not access sensitive local files (e.g., .ssh, .aws) or hardcode credentials.
[COMMAND_EXECUTION]: The skill does not execute arbitrary shell commands. It provides templates for SQL and Terraform but does not invoke them directly.
[EXTERNAL_DOWNLOADS]: No suspicious external downloads or remote script executions were identified. References to external entities (PostgreSQL, Redis, Terraform) are standard for the domain.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests user requirements to generate architectural designs (an inherent surface for indirect injection), it maintains a structured workflow and emphasizes security guardrails (like fail-closed policies). Standard LLM safety filters apply.
Ingestion points: User-provided SaaS requirements and database constraints.
Boundary markers: The skill uses a structured ASSESS -> STRATEGY -> DESIGN workflow to separate analysis from generation.
Capability inventory: Generates SQL, architectural documentation, and IaC templates; no direct system-level execution capabilities.
Sanitization: Not explicitly implemented in the instructions, but the output is intended for developer review.

shard