Teacher’s Edition
Module 12
Skills, Schedules, and Automation
Claude goes from a tool you open to a system that works on its own. Skills are reusable instruction sets. Schedules trigger them automatically. A weekly sales report that builds itself every Friday at 4 PM.
Charter Oak Strategic Partners · Claude Mastery Program · Version 1.0 · Confidential · Not for distribution to participants
Skills are a product feature in Claude, not a metaphor. A Skill is a folder of instructions, scripts, and resources that Claude loads dynamically when relevant to a task. When you ask Claude to do something, it reviews available Skills, loads the ones that apply, and follows their instructions. Skills improve consistency, speed, and quality on specialized tasks by giving Claude domain-specific procedures rather than relying on general knowledge.
The concept works through progressive disclosure: Claude reads the Skill’s metadata first (name and description), then loads the full instructions only if the Skill is relevant to the current task. This prevents context window overload when many Skills are available.
Anthropic Skills. Created and maintained by Anthropic. These include enhanced document creation for Excel, Word, PowerPoint, and PDF files. They are available to all users and Claude invokes them automatically when relevant. Participants have likely used these without realizing it — when Claude creates a formatted spreadsheet or a styled document, Anthropic Skills are at work.
Custom Skills. Created by individual users or organizations. These are the focus of this module. A custom Skill could enforce brand guidelines in every document, apply a specific analytical framework to data, automate a team’s reporting workflow, or ensure compliance requirements are met in every output. Custom Skills range from a few lines of instructions to multi-file packages with executable code.
Partner Skills. Professionally built Skills from partners like Notion, Figma, and Atlassian, available through the Skills Directory. These work with their respective MCP connectors to enable integrated workflows. For example, a Notion Skill combined with the Notion MCP connector lets Claude read and update Notion databases as part of a workflow.
Organization-provisioned Skills (Team and Enterprise). On Team and Enterprise plans, organization Owners can provision Skills for all users. Skills provisioned this way appear automatically in every team member’s Skills list and can be set as enabled or disabled by default. This allows companies to package standard operating procedures, brand guidelines, and compliance workflows as Skills that every employee’s Claude instance follows.
This distinction is critical for participants to understand. It will come up in questions:
Custom Instructions (Settings → Personalize) apply broadly to every conversation. They define how Claude communicates with you — tone, format preferences, background about your role. They are always on.
Projects provide static background knowledge that is loaded when you start a chat within that project. Upload documents, define project-specific instructions, and Claude uses them for every conversation in that project. The knowledge is always loaded when you are in the project.
Skills provide specialized procedures that activate dynamically when needed and work everywhere across Claude — in Chat, in Projects, and in Cowork. They are task-specific: Claude only loads a Skill when the current task is relevant. This makes Skills better for specialized workflows than custom instructions (which are always on) or projects (which require you to be inside the project).
Analogy for the room: Custom instructions are your preferences. Projects are your filing cabinets. Skills are your procedures. Preferences apply everywhere. Filing cabinets are organized by topic. Procedures activate when the right situation arises.
Opening: What Is a Skill? — 15 minutes
demo-data/module-12/sample-skill-file.md— Working SKILL.md with YAML frontmatter, instructions, and cron reference.“Let me show you what a Skill looks like under the hood.” Display the sample SKILL.md file on screen.
“Every Skill starts with a SKILL.md file. This is the file Claude reads. It has two parts: metadata at the top, and instructions in the body.”
Walk through the metadata: “The name field tells Claude what this Skill is called. Keep it under 64 characters. The description field tells Claude when to use this Skill. This is the most important field — Claude reads the description to decide whether to load the Skill for the current task. A vague description means Claude will not know when to activate it. A specific description means Claude loads it at exactly the right time.”
Walk through the body: “The markdown body contains the actual instructions. What the Skill does, step by step. What inputs it expects. What format the output should take. What constraints to follow. Think of it as a standard operating procedure that Claude executes instead of a human.”
“For Tier I and II graduates, this should feel familiar. A Skill is a system prompt with a purpose and a trigger.”
Walk through each section of a well-structured Skill:
YAML frontmatter (required):
---name: weekly-sales-reportdescription: Generates a formatted weekly sales summary from CRM export data. Use when processing sales CSVs or creating recurring revenue reports.---
Instructions (markdown body): What the Skill does, step by step. Include constraints (word count, format, audience), input expectations (file types, data shape), and output specifications (file name, format, where to save).
Reference files (optional): If the Skill needs supplementary information — a brand guidelines document, a list of approved terminology, an analytical framework — add it as a separate file (e.g., REFERENCE.md) in the Skill folder and reference it from SKILL.md. Claude loads reference files only when needed, keeping the initial context light.
Scripts (optional, advanced): For advanced Skills, attach executable code files. Claude can run these scripts as part of the Skill execution. Anthropic’s document Skills (Excel, PowerPoint, PDF) use scripts to create formatted files. Most custom Skills do not need scripts — markdown instructions are sufficient.
Live Demo: Build a Skill in Three Ways — 20 minutes
“The easiest way to create a Skill is to ask Claude to help you build one.”
Open a Claude chat (or Cowork). Type: “Help me create a Skill for processing meeting notes. It should take raw meeting notes as input and produce three things: a clean summary (five bullet points maximum), a table of action items with owner and deadline columns, and a draft follow-up email to attendees.”
Show Claude’s response: it will ask clarifying questions about format, tone, and constraints. Answer them. Claude produces a SKILL.md file. Review it with the room: “Look at the metadata. Is the description specific enough for Claude to know when to activate this? Look at the instructions. Are the steps clear enough that Claude will produce the same output every time?”
This method works for most business users. No file system knowledge required.
“For Skills that need reference files or scripts, you package them in a folder and upload as a ZIP.”
Show the folder structure on screen:
meeting-notes-skill/
SKILL.md
references/
email-template.md
“ZIP the folder. Go to Settings → Customize → Skills. Click ‘Add custom skill.’ Upload the ZIP. The Skill is now available in every Claude session.”
Demonstrate the upload. Show the Skill appearing in the Skills list. Show that it can be enabled or disabled.
“For organizations on Team or Enterprise plans, Owners can provision Skills for the entire team. The Skill appears in every team member’s Claude instance automatically. No individual setup required.”
Show the admin interface (if available) or describe it: “Organization Owners go to the admin panel, upload the Skill, and choose whether it is enabled or disabled by default. Every team member gets it. This is how you standardize workflows across a department or company.”
Concrete examples: “Your compliance team creates a Skill that checks every document for regulatory language. Your brand team creates a Skill that enforces voice guidelines in every piece of writing. Your finance team creates a Skill that formats every report according to the company template. All of these activate automatically when the relevant task comes up.”
The conversational method (Method 1) is where most participants should focus. Methods 2 and 3 are worth showing so participants understand the full spectrum, but the hands-on exercise should use Method 1 unless participants are technically comfortable with file management.
The AgentSkills Open Standard — 5 minutes
The Skill specification is published as an open standard at agentskills.io. Skills you create for Claude work across any AI platform or tool that adopts the standard. A reference Python SDK is available for developers implementing Skills support in their own platforms.
Why this matters for the room: “If you invest time building a library of Skills for your organization, that investment is portable. You are not locked into one vendor. The format is open.”
Scheduling in Cowork — 10 minutes
Cowork supports scheduled tasks that run automatically or on demand. There are two ways to create them:
From a Cowork session: Type /schedule in any Cowork conversation. Describe what the task should do, when it should run, and what folder it should use.
From the sidebar: Click “Scheduled” in the Cowork left sidebar to view, create, and manage scheduled tasks.
The critical constraint: Scheduled tasks only run while the computer is awake and Claude Desktop is open. This is fundamentally different from a server-based automation. If the laptop is closed at the scheduled time, the task does not run. State this clearly and repeatedly. Participants will assume it works like a cloud service.
A cron expression defines when a task runs. Five fields: minute, hour, day of month, month, day of week. Participants do not need to memorize the syntax. They need to recognize the six patterns that cover 90% of business use cases:
0 9 * * 1-5 — Every weekday at 9:00 AM
0 8 * * 1 — Every Monday at 8:00 AM
0 17 * * 5 — Every Friday at 5:00 PM
0 9 1 * * — First of every month at 9:00 AM
0 */4 * * 1-5 — Every 4 hours on weekdays
30 8 * * 1 — Every Monday at 8:30 AM
Show the reference table on screen. “You do not need to memorize this. You need to know it exists so you can look it up when you need it. Or ask Claude to write the cron expression for you — it is good at that.”
Open Cowork. Select a folder containing the weekly sales CSV. Type /schedule.
“Create a scheduled task that runs every Friday at 4 PM. Read the latest CRM export from this folder, calculate the week’s key metrics (total revenue, deals closed, pipeline value, win rate), compare to last week’s numbers, and save a formatted report as weekly-sales-report.md.”
Show the task appearing in the scheduled tasks list. Show that it can be paused, edited, run manually, or deleted. Run it manually once to demonstrate the output.
“Set once. Runs every week. The report appears in the folder every Friday afternoon without anyone opening Claude. The one caveat: your laptop needs to be open and Claude Desktop needs to be running at 4 PM on Friday.”
Group Exercise: Build, Schedule, Present — 30 minutes
“Pick a recurring task you do at least once a week. Something that takes 15 minutes or more and follows a consistent process.”
“Phase 1 (15 minutes): Build a Skill for it. Use Method 1 — ask Claude to help you create the SKILL.md. Define the trigger, the input, the process, and the output. Test it by running the Skill once with sample data.”
“Phase 2 (5 minutes): Write the cron expression. When should this run? Use the reference table. If you are unsure, ask Claude to write the expression for you.”
“Phase 3 (5 minutes): Calculate the ROI. How many minutes does this task take manually? How often do you do it? Multiply. That is your annual time savings.”
“Phase 4 (5 minutes): Present to your table. Two minutes each. What you built, when it runs, how much time it saves.”
Some participants will try to build a Skill that handles ten steps and five edge cases. Redirect them: “Build the 80% case first. Get it working. Handle edge cases later. A simple Skill that runs reliably is worth more than a sophisticated Skill that does not.”
Remind them of Anthropic’s own guidance: start with basic instructions in Markdown before adding complex scripts. Keep Skills focused — create separate Skills for different workflows rather than one large Skill that tries to do everything.
Some participants will say they do not have a recurring task. They do. They just do not think of it that way. Prompt them: “Do you send a weekly status update? Do you prepare for meetings? Do you review reports? Do you onboard new team members? Do you update a tracker?” Every knowledge worker has recurring tasks. The exercise is designed to surface them.
Debrief: The Math — 10 minutes
“Let us add it up. How many Skills did we build in this room?” Count them. “What is the total time savings per week?” Add it up on the board or projector.
“If fifteen people each built a Skill that saves two hours per week, that is thirty hours per week returned to the team. 1,560 hours per year. Hours that used to go to repetitive formatting, compilation, and data entry — now available for the work that actually requires your expertise, your judgment, your relationships.”
“Now multiply. One Skill saves two hours per week. Ten Skills save a full working day. The people who build five Skills in their first month typically build twenty in their first quarter. Each Skill you build teaches you to spot the next automation opportunity faster. The compounding effect is real — you get better at the work that matters because you spend less time on the work that does not.”
Before participants start building Skills for their teams, address security:
Review Skills before installing. Skills can include executable scripts and instructions that connect to external services. Read the SKILL.md before uploading it. Understand what it does and what access it requires.
Be cautious with third-party Skills. Skills from the Skills Directory are reviewed by partners, but custom Skills from unknown sources should be treated like any untrusted code. Review the instructions and scripts before enabling.
Code execution must be enabled. Skills that include scripts require code execution to be enabled in Claude’s settings (Settings → Capabilities). For organization-provisioned Skills on Team and Enterprise plans, admins control this centrally.
MCP connector awareness. Skills that reference external services (databases, APIs, file systems) may use MCP connectors. Participants should understand what data leaves their machine when a Skill activates a connector.
| Segment | Activity | Time |
|---|---|---|
| Opening | What Skills are, three types, how they differ from Projects/Instructions | 15 min |
| Demo | Three methods: conversational, upload, org-wide provisioning | 20 min |
| Standard | AgentSkills open standard and portability | 5 min |
| Scheduling | Cron expressions, Cowork scheduled tasks, live demo | 10 min |
| Exercise | Build a Skill, write the schedule, calculate ROI, present | 30 min |
| Debrief | Aggregate ROI calculation, security, next steps | 10 min |