AI Workflows and Automation skills
AI Workflows and Automation Skills-md for AI Agents
AI workflows and automation skills are SKILL.md files that help Claude, Codex and other AI agents plan, build and run AI at work. They cover choosing what to automate, writing prompts and skills, designing agents, using Claude Code for non-coding work, connecting tools with MCP, building Zapier, Make and n8n automations, processing documents, building knowledge assistants, testing quality, designing human review, securing data and controlling costs.
13 of 13 skills published, version 1.0.0, October 2026.
AI Workflow Audit
Audit how a person or team uses AI and automation: a task inventory, current AI use and risks, task scoring for frequency, time, repeatability, error cost, data sensitivity and judgement, the simplest pattern per task from assistant to agent, quick wins, guardrails, baselines and a 90-day roadmap.
Prompt Engineering for Work
Write and improve prompts for everyday work and team templates: success criteria, context and audience, structured inputs with XML tags, few-shot examples, rules and output formats, step-by-step reasoning, prompt chains, handling unknowns, testing on real cases, failure fixes and saving prompts as templates and skills.
Writing SKILL.md Files
Write and improve Agent Skills: SKILL.md anatomy and frontmatter, descriptions that trigger on real requests, focused jobs, principles, workflows, templates, worked examples, edge cases, handoffs, output formats and self-checks, progressive disclosure with supporting files, white-labelling, trigger tests and versioning.
AI Agent Design
Design AI agents and agentic workflows: deciding between workflows and agents, patterns from prompt chaining and routing to orchestrator-workers and evaluator loops, goals and stop rules, tool design, system prompts, context, guardrails and approvals by risk, testing on real tasks and monitoring.
Claude Code for Work
Use Claude Code for non-coding business work: project folders, a CLAUDE.md with goals and rules, skills and connectors, task briefs, plan review before changes, working on copies, checking outputs, safety rules and turning repeated tasks into reusable skills, for data clean-up, analysis, reports, documents, QA and content operations.
MCP Connectors
Plan, set up and govern MCP connectors for AI assistants and agents: MCP basics, mapping workflows to tools, choosing official or trusted servers, least-privilege permissions, OAuth and secret handling, read-only testing, usage and approval rules, prompt injection safeguards and a connector inventory with reviews.
No-Code Automation
Design, build and maintain no-code automations with AI steps in Zapier, Make, n8n or native tools: process mapping, platform choice, triggers, early filters, lookups, AI steps with JSON outputs, routers, approvals, error handling, testing, monitoring, cost control and documentation.
Document Processing with AI
Extract and summarise documents with AI: field schemas, document preparation, extraction prompts with JSON output and no guessing, automatic validation of totals, dates and duplicates, confidence-based review routing, accuracy testing, integration and monitoring for invoices, contracts, forms, CVs and reports.
RAG and Knowledge Bases
Design knowledge bases and retrieval-augmented generation for teams: real question inventories, source audits and clean-up, choosing between project files, connectors, enterprise search and custom RAG, structure and metadata, answer prompts that cite or decline, permissions, evaluation scorecards and maintenance.
AI Evals and Testing
Design evals for prompts, skills, workflows and agents: success criteria, test sets with real, edge, adversarial and past-failure cases, grading with code checks, reference answers, rubrics, pairwise comparison and calibrated model graders, baselines, regression testing, ship decisions and production monitoring.
Human-in-the-Loop Design
Design human review for AI workflows: action risk classification, review levels from spot checks to approve-before-action and human decisions, confidence routing, review screens with source and checklists, service levels, reviewer training, feedback loops, rubber-stamping checks and evidence-based autonomy.
AI Security and Privacy
Set practical AI security and privacy rules: data classification and approved tools, OWASP LLM risks, prompt injection defences, least privilege for agents and connectors, secrets handling, insecure outputs, logging, vendor checks, a one-page staff guide and incident response.
AI Cost and Model Selection
Choose AI models and control costs: model tiers by task, quality bars from evals, cost per task from tokens, per-step model choice in chains and agents, shorter prompts, prompt caching and batching, agent limits, tool and seat consolidation and monthly reviews of spend against value.
No skill matches that search. Clear it or choose All.
What these skills do
Most teams adopt AI in scattered ways: a few people paste work into chat tools, someone builds an automation nobody else understands, and pilots stall because nobody measured the before and after. Results stay small while risks such as data leaks and prompt injection stay hidden.
These skills give AI work a method. They start from real tasks and pick the simplest pattern that works, following Anthropic's advice to begin with simple workflows before agents. Prompts get context, examples and output formats; repeated tasks become SKILL.md files; agents get precise tools, stop rules and approvals; connectors get least-privilege access; and every workflow is tested with evals, reviewed by people where risk lives and costed per task.
Prompts, agents, automations and guardrails
The set covers strategy, prompting, skills, agents, Claude Code, MCP connectors, no-code automation, documents, knowledge bases, evals, human review, security and cost.
Where to start: tasks before tools
Start with ai-workflow-audit. List the tasks your team repeats, score them for frequency, time, repeatability, error cost, data sensitivity and judgement, and choose the simplest pattern for each: a prompt template, an automation, an AI workflow or, only where needed, an agent. Set guardrails and baselines before building.
Then use prompt-engineering and skill-md-authoring for quick wins: most teams save hours a week from a few well-built prompts and skills before any automation. Move on to no-code-automation, document-processing or ai-agent-design for bigger workflows, and run ai-evals-testing before every launch.
Security, human review and cost are part of every build, not a later phase: ai-security-privacy, human-in-the-loop and ai-cost-model-selection keep workflows safe, trusted and affordable as they scale.
A typical first quarter looks like this. In the first two weeks, run the audit and ship two or three prompt templates or skills for the most repeated tasks, measuring time before and after. In weeks three to six, build one automation or document workflow with structured AI steps, approvals for risky actions and an eval set of real cases. In weeks seven to twelve, connect tools through MCP where useful, pilot an agent only if a task truly needs one and review costs per task. By the end of the quarter, most teams have measurable time savings, a short list of trusted workflows with owners and documentation, and clear rules on data, review and spend that make the next round of automation faster and safer.
Every skill also works on its own when you only need one part.
The 13 skills and the sources behind them
All thirteen skills in this set are live.
| Skill | What it does | Status |
|---|---|---|
ai-workflow-audit | Find tasks to automate and choose the simplest pattern | Live |
prompt-engineering | Prompts with context, examples, formats and tests | Live |
skill-md-authoring | Write SKILL.md files that trigger and perform | Live |
ai-agent-design | Agents with good tools, stop rules and guardrails | Live |
claude-code-for-work | Claude Code for spreadsheets, reports and documents | Live |
mcp-connectors | Connect AI to your tools with least privilege | Live |
no-code-automation | Zapier, Make and n8n flows with AI steps | Live |
document-processing | Extract and validate data from documents | Live |
rag-knowledge-bases | Assistants that answer from your documents with citations | Live |
ai-evals-testing | Test quality before and after every change | Live |
human-in-the-loop | Review where risk lives, without rubber-stamping | Live |
ai-security-privacy | Data rules, prompt injection defences and secrets | Live |
ai-cost-model-selection | The right model for each step at the right cost | Live |
The sources and rules behind these skills
| Source or rule | How the skills use it |
|---|---|
| Anthropic, Building effective agents | Start simple; workflows before agents; proven agent patterns |
| Anthropic prompt engineering guidance | Clear instructions, examples, XML structure, reasoning and chaining |
| Agent Skills open standard | Descriptions that trigger, focused SKILL.md files, progressive disclosure |
| Model Context Protocol | Connectors with least privilege and approvals |
| OWASP Top 10 for LLM applications | Prompt injection, data disclosure and excessive agency defences |
| Evals before iteration | Success criteria, test sets and regression testing |
Example prompts
Each skill triggers on everyday wording. These prompts are good starting points:
- Audit: “Which of our team's tasks should we automate with AI?”
- Prompts: “Fix this prompt; the answers are too generic.”
- Skills: “Turn our onboarding checklist into a SKILL.md.”
- Agents: “Design an agent that researches 30 accounts a week.”
- Claude Code: “Set up Claude Code to build our monthly report.”
- Automation: “Route website enquiries with n8n and an AI step.”
- Evals: “Test this prompt on 50 real tickets before we launch.”
What the agent needs from you
Good AI workflow design depends on a few inputs:
- The task: what it is, who does it, how often and what good looks like.
- Real examples: inputs and ideal outputs, including hard cases.
- Tools and data: the apps involved and the data they hold.
- Rules: data sensitivity, approvals and company policies.
- Measures: time, quality and cost before you start.
If something is missing, the skills ask only what changes the design and start with safe defaults.
What an agent cannot do, and myths to ignore
An agent can design, draft, build and test, but it cannot decide your risk appetite, guarantee correctness or replace legal and security review for sensitive data. People stay accountable for what AI workflows do.
| Myth | What works better |
|---|---|
| Agents are always better | Simple workflows are cheaper and more reliable when steps are known |
| A good prompt works forever | Re-test with evals when inputs, prompts or models change |
| Human approval makes it safe | Only if reviewers see the source and are not rubber-stamping |
| The biggest model is safest | Choose the model that meets your quality bar per step |
| AI can read any email or page safely | External content can carry prompt injection; treat it as data |
Frequently asked questions
How do I start using AI at work?
List repeated tasks, score them, pick the simplest pattern, set guardrails and measure before and after.
When should I build an AI agent?
When steps vary by case and need several tools; otherwise a workflow or prompt is better.
What is a SKILL.md file?
A folder-based instruction file with a name and description that teaches an AI agent how to do a task well.
What is MCP?
The Model Context Protocol, an open standard for connecting AI applications to tools and data.
How do I test AI quality?
Define success criteria, build a test set of real and edge cases and compare versions with evals.
How do I protect against prompt injection?
Treat external content as data, limit tools, require approvals and test with adversarial cases.
Which agents can use these skills?
Claude, Claude Code, Codex, Cursor, Gemini CLI and other tools that support the Agent Skills format.
Are these skills free?
Yes, under CC BY 4.0, as long as you credit Afeez Adebayo and link to the source page.
Sources and more skills
- Anthropic: Building effective agents
- Anthropic prompt engineering overview
- Agent Skills specification
- Model Context Protocol
- OWASP Top 10 for LLM Applications
Checked on 14 October 2026. See all Productivity skills and the GTM Engineering skills.
