@Victor Vincent I agree with you. Skills are not a one-size-fits-all solution, they must be niche or business specific. Skills developed with SOP for a vertical will work for the same category, with exceptions and custom needs.

@Mikael Wällstedt What I mean may not be applicable to all use cases, but it works extremely well as an internal tooling. Let me try to describe what I mean. Scenario: HR Recruitment AI Tools GTM model (ATS): no dashboard, no SaaS, tools are available wherever the user is, technically there is a dashboard, but only for onboarding and maintenance. Example: Use Case (Candidate Screening) 1. Recruiter remains on his/her ATS dashboard (our AI widget is embedded in their dashboard). 2. Recruiters get their candidate screening done in real time using real data, without getting into the technical details of how our AI tools work. 3. Various AI Tools (Agents/Workflows) handle different tasks based on the recruiter's needs and our offerings. Now we'll look at the Sideloader Companion skill. It's an extended version for the lightweight task. For example, candidate screening Skill 1. HR recruiters receive applications from various channels, including multiple mailboxes, datasheets, and print (OCR) resources. 2. The recruiter does not need to log into their ATS to upload or scan, it can be done with the help of the Skill (LLM) and produce the same results. Output formats include datasheets, CSV, PDF, and reports, which can be synced with the tenant account's dashboard. Which I mentioned earlier. The Skill can be used as a standalone workflow or as an internal tool for the tenant, which is what I mean by "sideloader."

@Fausto Rodriguez My take on cold accounts is to avoid paid promotions unless it is a product or service. Instead, allow your account's algorithm to warm up organically at this point, paid promotions will bring in generic audience, which is noise rather than prospective leads. If you're a founder or authority, make text-only posts from your LinkedIn and X accounts, while creative posts once or twice a week can come from personal or company pages on Facebook, Instagram, X, and LinkedIn. TikTok is a different breed, and it is not an appropriate source for your content tone and language.

@Jorge Martínez According to the workflow, the initial trigger begins with the Telegram trigger and proceeds through an HTTP request to retrieve the payment link, which is then sent to the user via the Merge node as a response. When a user clicks on a payment link, it is routed to a Webhook, which checks for Success or Failed before returning to the Merge node and Telegram response. If I am understanding correctly, there is a problem here. When you use a Webhook trigger, no Telegram user data is included. To resolve it there are many ways to do it. The important aspect is to keep the user chat ID. Solutions: 1. Save and append the user chat ID to the Payment Link so that Webhook can identify and retrieve it when triggered. 2. Save it in your workflow so that it can be passed to the Telegram response. 3. If you know the chat ID and it is fixed (for example, a Telegram bot), you can hardcode it into the Telegram response. I hope I was able to understand your situation.

@Jorge Martínez For option 2, you will need to store the data into a table with fields like chat ID and payment ID or any unique identifier. When the webhook is triggered, retrieve data with a unique identifier, such as a payment ID that matches the webhook payment ID, get chat ID from the table associated with that payment ID, and pass it to your Telegram response.

As @Kenan Kurtuluş said, you will need to add the account as a test user under the OAuth consent screen: From the dashboard, navigate to APIs & Services and click OAuth consent screen. Add a test user and, if your API has restrictive access, enter your n8n domain to grant access.

What tools, platforms, or frameworks are you planning to use? I might be able to help.

@Ryan Nickel Although it has limitations, widgets can be added to a client's website to provide interactive tools. If you want a completely customized user interface, I would recommend looking into Pydentic AI, which is a very strong framework. GoHighLevel will be a better option, but all three can be accomplished with n8n.