What Are Skills?
Skills are saved instructions that teach your AI a specific workflow, exactly the way your team runs it.
Every time you repeat a task, you re-explain the same context: which tools to check, what format the output needs, what rules to follow. A skill captures that once. After that, one sentence is enough.
How Skills Work
A skill is a document that defines a workflow from start to finish.
It tells the AI what phrases should trigger it, which connectors to call and in what order, how to make decisions along the way, and what the final output should look like. The AI reads it at the start of a relevant task and follows it as a blueprint.
Creating and Editing Skills
The easiest way to create a skill is to describe what you want in plain language. Tell the AI the outcome you're after, not the step-by-step process.
A session typically looks like this:
You: I want the AI to handle our weekly support ticket summary.
AI: Got it. A few things to nail down:
What counts as urgent?
Which channels should be included?
Where should the summary go, email or Slack?
You: Urgent means open 48+ hours or tagged billing.
All channels. Post to #support-ops.
AI: [Writes the complete skill document]
Once saved, every future session picks it up. No re-explaining.
"I want a weekly support digest that flags anything urgent and posts to Slack" gets you further than "first query Zendesk, then filter by tag, then format as a list..." Over-specifying the steps tends to confuse the output. Describe the goal and let the AI figure out the implementation.
Didn't like the response quality? Give feedback to the agent for further updates of the same skill.
Skills are plain-text documents, so you can also write or edit them manually if you prefer that level of control.
Most people start with a conversation and then tweak the result.
Importing Skills
You can also import skills from other agentic tools you already use (like Notion AI, Claude, or any system that exports agent instructions) and run them inside TextCortex.
And you can export skills you've built here to use elsewhere. If your team has already invested in building workflow instructions in another tool, you don't start from scratch.
Either way, the goal is the same: describe what you want done, save it once, and never explain it again.
What a Skill Contains
Trigger phrases are the sentences that activate the skill. A project management skill might fire on "create a brief for X" or "set up tasks for this sprint". A support skill might activate on "summarize open tickets" or "draft a reply to this complaint".
Connector calls define which tools get used and when. A customer support skill might pull open tickets from Zendesk, check order history from Shopify, then draft a reply. A project skill might create tasks in Asana, log notes in Notion, and post an update to Slack, in that order.
Decision logic encodes the judgment calls your team makes repeatedly. A support skill might define that any ticket open for more than 48 hours with no reply gets flagged as urgent. An email marketing skill might define that segments under 500 contacts always get a plain-text version.
Output format locks in what the deliverable looks like every time. A weekly report skill always produces the same structure. A lead list skill always includes the same columns. No reformatting, no reminders.
Hard rules are the constraints that matter most: never close a support ticket without a resolution note, never add someone to a drip sequence if they're marked as churned, never push a task to the sprint board without an owner assigned.
Examples by Team
Customer Support
A support skill connects Zendesk and Shopify. You say "summarize urgent open tickets from this week" and it pulls all unresolved tickets, checks order status for each customer, groups them by issue type, and returns a prioritized list with suggested replies.
The skill defines what "urgent" means (open 48+ hours, or tagged billing or shipping). You defined that once. Now it just applies.
Project Management
A project kickoff skill connects Notion, Asana, and Slack. You say "set up the Q3 website redesign project" and it creates a brief in Notion, builds the task structure in Asana with assignees and due dates, and posts a summary to the right Slack channel.
3 tools, 1 sentence.
Email Marketing
A campaign skill connects Klaviyo and your CRM. You say "draft a re-engagement sequence for customers who went quiet in the last 6 months" and it pulls the right segment, checks contact history, and drafts a 3-email sequence with timing, subject lines, and copy.
The skill carries your brand rules: tone, banned phrases, which offers apply to which segments. Every draft comes out on-voice.
Sales
A prospecting skill connects a contact database and your CRM. You say "find IT decision-makers at mid-market manufacturing companies in Germany" and it searches for matching accounts, enriches the shortlist, cross-checks the CRM to exclude existing customers and active deals, and returns a formatted lead list with ownership flags.
Nothing lands on the list without passing the CRM check first.
Engineering
A bug triage skill connects Jira and Slack. You say "summarize this week open bugs by severity" and it pulls all open issues, sorts them by priority, flags anything unassigned for more than 3 days, and posts a digest to the engineering channel.
More Specialized Skills
Once your team is comfortable with the basics, skills can go deeper.
A content research skill can scan your entire blog CMS, run keyword research across multiple query types, benchmark against competitors, and produce a full briefing document with article ideas, difficulty scores, and traffic estimates. An SEO audit skill can pull domain metrics, surface keywords ranking just outside the top 10, identify pages losing traffic, and generate a downloadable report.
A prospecting skill built for a sales team can apply intent scoring, flag contacts who recently changed roles (a strong buying signal), and cross-reference 2 systems before any name makes it onto the list. The structure is the same as a simple skill. The connectors and decision logic are just more layered.
When to Create a Skill
If you find yourself typing the same context block at the start of a session, that block belongs in a skill.
A task is ready to become one when you have done it 2–3 times, the process is consistent, and the output format actually matters. If a wrong output causes rework, that is a skill waiting to be written.