6 Comments
User's avatar
Pratik Mehta's avatar

Hi, can these skills be used to analyze newsletters across substack or for applicable to personal newsletters only?

Alejandro Aboy's avatar

As long as you provide publication URLs, you can use the Substack Author MCP (or the MCP + the skills) - there's no "newsletter Discovery" in place but I might add something on that end.

Suchi Raman's avatar

Thanks for a detailed article...so the Skills is loaded only when the agent recognised the context ...what happens if it encounters a scenario not in the specific context

Alejandro Aboy's avatar

Thanks for the comment! That's what observability is for. We can see those unexpected scenarios and prepare the Agent skills to cover them, the system prompt to consider them or just let them fail since they might out of scope and not relevant.

Pawel Jozefiak's avatar

The "skills tell the agent how and when" framing is exactly what clicked for me too. Modular instruction files keep context clean - the alternative is a system prompt that becomes unwieldy after the 10th tool added.

What I found missing: visibility into WHICH skills fired during a session and why. Ended up building a dashboard just to understand my own agent's behavior: https://thoughts.jock.pl/p/wiz-1-5-ai-agent-dashboard-native-app-2026

The Substack content analysis angle is interesting. Did you find it handles cross-post topic tracking, or only individual post performance?

Alejandro Aboy's avatar

Thanks for your comment Pawel. The Substack Author MCP can understand if you cross post articles and then the AI can extract topics if you need, also individual post performance can be retrieved with another tool.