How to Install an AI Workflow Pack (Step-by-Step)
Learn how to install OutcomeKit workflow packs into Claude Code, Cursor, Hermes Agent, OpenClaw, and any Agent Skills-compatible agent.
What is a workflow pack?
An OutcomeKit workflow pack is a folder of structured files that give an AI agent the knowledge and decision logic to perform a specific business task. Each pack follows the open Agent Skills format (agentskills.io), which means it works across any compatible agent — Claude Code, Cursor, Hermes Agent, OpenClaw, VS Code Copilot, and many more.
The core file is SKILL.md, which contains YAML frontmatter (name, description, compatibility, metadata) followed by the full operating instructions: purpose, workflow steps, decision rules, edge cases, output requirements, and failure handling. Beyond SKILL.md, packs include input/output schemas, worked examples, test cases, and a QA checklist.
Supported agents
OutcomeKit packs work with any agent that supports the Agent Skills open standard. Here are the most common agents and where they expect skills to be installed:
Claude Code: Place the skill folder in ~/.claude/skills/ for global access, or in your project's .claude/skills/ directory for project-scoped skills. Claude Code discovers SKILL.md files automatically.
Cursor: Place the skill folder in .cursor/skills/ at your project root. Cursor loads skills from this directory when you open the project.
Hermes Agent: Place the skill folder in ~/.hermes/skills/ organized by category (e.g. ~/.hermes/skills/sales/qualify-inbound-leads/). You can also install from the Hermes skills registry using hermes skills install. Hermes is fully compatible with the agentskills.io format.
OpenClaw: Place the skill folder in your workspace's skills/ directory. OpenClaw also checks ~/.openclaw/skills/ for global skills. You can install skills from ClawHub using openclaw skills install.
VS Code Copilot, GitHub Copilot, Gemini CLI, Roo Code, and 20+ other agents also support the Agent Skills format. Check your specific agent's documentation for the exact skills directory path.
What's inside a pack
When you download and extract a pack ZIP from OutcomeKit, you'll find a structured folder that looks like this:
SKILL.md — The main file. Contains YAML frontmatter with metadata (name, description, compatibility, version) and the full operating instructions your agent follows.
manifest.json — Machine-readable metadata: category, version, setup time, compatibility list, required inputs.
input-schema.json and output-schema.json — JSON Schema contracts that define what the skill expects as input and what it produces as output. Agents use these to validate their work.
examples/ — Worked examples showing good output (3 cases), bad-fit scenarios (1 case), and edge cases (1 case). These give the agent concrete reference points.
tests/ — An evaluation test set with pass/fail criteria. Use this to verify the agent is producing correct output.
checklists/ — A QA checklist for human review of the agent's output.
CHANGELOG.md — Version history for the pack.
Tips for getting the best results
Keep the skill folder name exactly as downloaded. The folder name must match the name field in the SKILL.md frontmatter for the agent to discover it correctly.
Don't edit SKILL.md unless you know what you're doing. The decision rules, edge cases, and constraints are carefully designed to produce reliable output. If you modify them, the test cases may no longer pass.
Use the examples as a reference. If your agent's output doesn't match the quality of the examples in the examples/ folder, something may be wrong with how the skill was loaded or invoked.
Run the test cases after installation. The tests/ folder contains an evaluation set. Feed the test inputs to your agent and compare against the expected outputs to verify the skill is working correctly.
Start with a free pack to test the workflow. Meeting Prep Brief and Weekly Metrics Summary are free packs that let you verify the install process before purchasing paid packs.
Step-by-step
- 01
Download the pack from your dashboard
After purchasing or acquiring a free pack, go to your OutcomeKit dashboard, open the pack, and click 'Download all as ZIP'. Extract the ZIP — you'll see folders like examples/, tests/, and checklists/ along with the core SKILL.md, manifest.json, and schema files.
- 02
Find your agent's skills directory
Each agent stores skills in a different location. Claude Code: ~/.claude/skills/ or your project's .claude/skills/ directory. Cursor: .cursor/skills/ in your project root. Hermes Agent: ~/.hermes/skills/ (organized by category). OpenClaw: your workspace's skills/ directory, or ~/.openclaw/skills/ for global access. Any Agent Skills-compatible agent: check your agent's docs for its skills directory — the SKILL.md format is an open standard supported by 30+ agents.
- 03
Copy the pack folder into place
Move or copy the extracted skill folder (e.g. qualify-inbound-leads/) into your agent's skills directory. Keep the folder name as-is — it must match the name field in SKILL.md. The agent will discover the skill automatically on next startup or session.
- 04
Verify the skill loaded
Most agents show installed skills on startup or via a command. In Claude Code, your skills appear in the system context. In Hermes Agent, run /skills to see the list. In OpenClaw, check your workspace skills. The agent reads the SKILL.md frontmatter first (name + description), then loads the full instructions when you activate the skill.
- 05
Use the skill in your workflow
Invoke the skill by referencing its name or describing the task it handles. The agent will load the full SKILL.md instructions, follow the decision rules, and use the input/output schemas to structure its work. Examples in the examples/ folder show what good output looks like.
Frequently asked questions
Do I need to install packs differently for each agent?
The pack format is the same everywhere — it follows the open Agent Skills standard (agentskills.io). The only difference is where you put the folder. Claude Code, Cursor, Hermes Agent, OpenClaw, VS Code Copilot, and 25+ other agents all read the same SKILL.md format.
What if my agent doesn't support the Agent Skills format yet?
You can still use the pack manually. Open SKILL.md and paste the instructions into your agent's system prompt or project instructions. The schemas, examples, and test cases work regardless of how you deliver them to your agent.
Can I install the same pack into multiple agents?
Yes. Copy the pack folder into each agent's skills directory. The files are identical — no conversion or reformatting needed.
What's in a pack besides the SKILL.md?
Verified packs include manifest.json (metadata), input and output schemas (JSON Schema contracts), worked examples (good, bad, and edge cases), an evaluation test set, a QA checklist, and a changelog. These give your agent everything it needs to run the workflow reliably.
Related packs
Ready to put this into practice? These workflow packs give you the instructions, schemas, examples, and tests to get started.
Keep reading
How to Automate Lead Qualification with AI Agents
Learn how AI workflow packs qualify inbound leads faster than manual review. Step-by-step guide for founders and small sales teams ready to stop losing deals.
OpsWeekly Ops on Autopilot: AI Workflows for Recurring Tasks
How founders and small teams automate weekly recurring tasks with AI workflow packs. Covers metrics summaries, meeting prep, founder updates, and building your ops stack.