Stop Prompting. Start Building.
Skills files give your agent expert-level tool knowledge — no prompt engineering, no trial and error.
A Skills file (SKILL.md) is a plain markdown document that encodes expert-level tool knowledge into your agent — the workflows, guardrails, and output formats condensed from hours of trial and error. No prompt engineering. No code. Works with any AI platform that supports project context or MCP.
How Skills Files Work
Three steps from download to a working, reliable agent — on any platform.
1) Download
Choose a free Skills file below. Enter your email and the .md file lands in your inbox with a platform-agnostic setup guide.
Output: SKILL.md + setup guide
2) Add to Your Agent
Upload the SKILL.md as context in your AI assistant's project or system prompt. Connect the matching MCP server listed in the file.
Output: agent with expert tool knowledge
3) Customise
Edit the Skills file to match your workflow — your tone, your guardrails, your team's conventions. It's just markdown.
Output: Skills file tailored to your stack
Learn how to use skills files well
These guides explain how skills files differ from prompt engineering, how MCP workflows work in practice, and how teams adopt them.
Learn how to use Skills files to make AI workflows repeatable: add the Skill as context, connect the MCP tool, run the workflow, and improve the Skill over time.
Compare skills files with one-off prompt engineering and see why reusable workflow context produces more reliable agent behavior across real teams.
Explore the most useful MCP-connected skill patterns for marketing, product, operations, and knowledge workflows, plus how to choose the right first skill.
What Is a Skills File?
Access without expertise is just noise. Skills files encode the expert knowledge so your agent stops guessing and starts performing.
The Problem
When you connect a Gmail MCP server to your AI agent, you've given it access. But access isn't expertise. Your agent doesn't know to read the full thread before drafting. It doesn't know to never auto-send.
The Solution
A SKILL.md file tells your agent exactly how to use each tool — step-by-step workflows, guardrails, output formats. Everything learned from trial and error, condensed into one file.
What's Inside
- Trigger conditions — when to use this skill
- Step-by-step workflows per use case
- MCP tool schemas + known gotchas
- Output format templates
- Guardrails — what the agent must never do
See what a Skills file looks like
--- name: gmail-assistant mcp_servers: ["Gmail MCP"] version: "1.0" --- ## When to trigger this skill - "check my inbox", "draft a reply to", "summarize this thread" ## Workflow 1: Inbox triage 1. list_emails(max_results=20, query="is:unread") 2. Classify each as: ACTION_NEEDED | FYI | NEWSLETTER 3. Return structured triage — 🔴 Action, 🟡 FYI, ⚪ Auto 4. Never mark as read without explicit user instruction ## Guardrails - NEVER call send_email() without explicit confirmation - NEVER bulk-archive without showing match list first - Always read the full thread before drafting a reply
Download Skills Files
Free skills are email-gated — we send the file directly to your inbox. Pro and Enterprise unlock the full library.
Instructs your agent to browse, extract, and return clean structured data from any public page — product listings, pricing tables, news articles.
Connects to Gmail MCP to summarize long threads, draft replies in your voice, and surface action items — without ever leaving your agent.
Pairs with Fetch MCP to crawl a URL, score on-page elements, and return a prioritized fix list your team can action immediately.
Integrates with Google Calendar MCP to find open slots, draft agendas, block focus time, and send invites — all via natural language.
Pulls from web search + your CRM MCP to enrich contact records with company size, funding, tech stack, and recent news before every call.
Builds a retrieval pipeline over Google Drive or Notion MCP so your agent answers questions with grounded citations — not hallucinations.
Breaks features into user stories with acceptance criteria, estimates in Fibonacci points, fills a sprint to your velocity, and generates standup/review/retro ceremony templates.
Researches competitive context, structures problem statements, writes MoSCoW requirements with SMART success metrics, and saves the finished PRD to Notion — feature-level or full initiative.
Reads merged PRs between two git tags, categorizes every change, and generates user-facing changelog, technical release notes, exec summary, email draft, and social snippet simultaneously.
Parses meeting notes or transcripts to extract decisions, action items, and open questions. Creates Jira tickets for every action item and writes a structured summary to Notion — in one pass.
Transforms bullet points or raw data into a Marp markdown presentation — exec update, roadmap reveal, QBR, product launch, or design review. Renders to PDF with one command.
Pulls backlog issues from Linear or Jira, walks through scoring, ranks by framework, flags stakeholder conflicts and quick wins, updates issues with scores, and outputs a shareable priority proposal.
Runs funnel analysis, retention cohorts, and metric drop investigations on PostHog or Mixpanel data. Returns narrative reports with WoW context, anomaly flags, and root cause hypotheses — not just raw numbers.
Creates and manages Linear cycles, triages the backlog, generates cycle health reports in three formats (full report, Slack summary, standup note), and flags blockers and over-assigned team members.
Searches the entire SharePoint tenant, organizes document libraries with a consistent taxonomy, creates structured wiki pages and team sites, and flags stale or misplaced files.
Applies Jobs-to-be-Done coding and affinity mapping to raw interview notes. Processes up to 20 interviews into a themed insight report, opportunity statements, HMW questions, and a shareable team readout slide.
Computes time-weighted OKR confidence scores from Linear and Notion data, tracks week-over-week trajectory, generates exec summaries and team updates, and facilitates mid-quarter pivot analysis with structured options.
Codes and affinity-maps NPS comments, support tickets, app reviews, and interview notes. Scores themes by frequency and business impact, surfaces churn signals, ranks feature requests, and includes competitive context.
Orchestrates web scraping, news monitoring, and structured analysis to deliver a weekly intel brief on up to 10 competitors — ready for Slack or email.
Connects to GitHub MCP to review PRs against your coding standards, flag security issues, suggest refactors, and post structured review comments.
Simple Pricing
Start free. Upgrade when your agent needs more power.
Free
Forever free — email required
- 3 Skills files (Web Scraper, Gmail, SEO)
- Setup guide with every download
- Weekly Skills newsletter
- Video walkthroughs coming soon
Pro
Full Skills library
- All Free skills +
- Calendar Planner, CRM Enricher, Doc Q&A
- Priority update drops
- Slack community access
- Email support
Enterprise
Bespoke skills for your stack
- All Pro skills +
- Custom Skills for your MCP setup
- Competitor Intel + Code Reviewer
- Dedicated onboarding call
- SLA + private Slack
Need a Custom Skills File?
We write bespoke Skills files for your exact MCP stack — Notion, Jira, Salesforce, your internal APIs. Delivered in 5 business days with an eval harness included.
Request a Custom Skill →MCP Servers & Context7
Need deeper integration work? We implement production-grade MCP servers and Context7 pipelines so your agents can reliably answer questions and take actions without guesswork.
MCP Server Setup
Expose approved tools (search, DB, docs, APIs) with guardrails, auth, and observability.
- Spec & scopes, auth strategy, rate-limits
- Tool definitions + eval harness
- Logging, red-teaming, rollback plan
Context7 Pipelines
Build retrieval flows that stay fresh and grounded (chunk, rank, dedupe, enrich, verify).
- Corpus design & data contracts
- Evaluator prompts & regression tests
- Latency + cost tuning, caching
How We Deliver
- Discover: scope + data map + success metrics
- Implement: MCP + pipelines + regression tests
- Harden: guardrails + monitoring + runbook
What is MCP & Context7?
MCP (Model Context Protocol) standardizes how agents access tools and data through a secure, observable server interface.
Context7 is a pragmatic framework for seven-layer context pipelines (ingest → normalize → chunk → rank → enrich → verify → deliver) that keep answers accurate and fast.