The Agent Skills Directory

Interactive Discovery Model: The skill relies on the AskUserQuestion tool for primary interactions, ensuring that major configuration decisions are mediated by the user rather than executed autonomously.
Project Analysis Surface: The skill uses Grep, Glob, and Read to scan the local codebase for existing Connect patterns. This allows it to pre-fill discovery questions based on the developer's current progress, enhancing utility without introducing privilege concerns.
Business Research Logic: It utilizes a Task tool to invoke a subagent for researching company URLs and business descriptions. This is a standard integration of business context into the recommendation workflow and does not involve unauthorized data access.
Internal Validation Matrix: The skill references a comprehensive suite of internal markdown files (e.g., compatibility-matrix.md, decision-matrix.md) to validate recommendations against known Stripe integration best practices, helping to avoid misconfigurations like liability-model mismatches.
Indirect Prompt Injection (Surface Analysis): The skill processes untrusted input such as company URLs and business descriptions which are then passed to a subagent. While this creates a theoretical attack surface for influencing the agent's logic, the interactive nature and the output format (markdown recommendation) significantly limit the risk of malicious exploitation.
Credential Safety: The skill contains no hardcoded API keys or secrets; all configuration is handled through logical dimensions and official documentation references (e.g., stripe.com/pricing).

connect-recommend