LotsAgent
You have an agent that handles lead qualification. It works. Now you want the same qualification logic in your support triage agent. Do you rebuild it from scratch?
If you said yes, you're doing it the hard way. Agent skills exist so you don't have to.
The Problem: Rebuilding Config for Every Agent
When you have one agent, configuration is simple. You set up memory, add tools, write instructions. It takes a few minutes. But when you scale to three, five, or ten agents — each with overlapping needs — you face a choice: duplicate config across agents (brittle and wasteful) or refactor into shared modules.
Skills are that shared module. They are the difference between manually copying instructions across five agents and updating one skill that every agent inherits.
What Agent Skills Actually Are
Skills are reusable packages of knowledge and workflow logic that you assign to agents. Think of them as plug-in capabilities.
| Concept | What It Is | Example |
|---|---|---|
| Skill | Reusable knowledge + workflow module | A "lead qualification" skill that scores leads using your criteria |
| Tool | External action via API | Gmail send, Slack post, GitHub create issue |
| Prompt | Instruction text in the system prompt | "Always respond in French" |
Skills are different from tools (which connect to external services) and prompts (which are static instructions). A skill packages both knowledge and behavior: it knows your lead scoring criteria and knows the workflow to apply them.
A skill is the difference between telling an agent what to do and giving it a capability it can execute.
Building a Skill via the Agent Builder
LotsAgent's Agent Builder creates skills from a conversation. You don't write code or configure YAML files.
Here's how it works — using a practical example:
Step 1: Describe the skill you need
In the Agent Builder, describe what you want:
"Create a skill that qualifies inbound leads. It should check the lead's company size, industry, and budget against our ideal customer profile. If it matches at least 3 of 5 criteria, mark it as qualified and notify the sales channel. If it matches fewer than 3, send a nurture email sequence."
Step 2: The builder configures it
The Agent Builder interprets your description and wires up:
- The knowledge: your ICP criteria, scoring thresholds
- The workflow: check → score → route to qualified or nurture
- The tools it needs: access to your CRM, email, and Slack channels
Step 3: Review before it's live
Nothing runs automatically. The builder presents the skill configuration. You review the criteria, the workflow steps, and which tools it needs. Approve, edit, or reject.
Step 4: Save and apply
The skill is saved once. From there, you assign it to any agent with a few clicks.
Applying Skills Across Agents
This is where the reusability pays off.
Before skills:
- Agent A: lead qualification (configured with instructions, criteria, and tools manually)
- Agent B: support triage (doesn't have lead qualification — needs separate setup)
- Agent C: account management (needs lead qualification — another separate setup)
- Total: three configurations to maintain, each slightly different
With skills:
- Create one "lead qualification" skill
- Assign it to Agent A, Agent B, and Agent C
- All three agents know your qualification criteria and workflow
- Update the skill once; all agents inherit the change
The source brief describes this as: "Build once, apply to many."
Managing Skills Over Time
Skills aren't static. As your workflows evolve, so should your skills.
Versioning — When you update a skill, agents that use it get the updated version. You don't need to reconfigure each agent.
Selective assignment — Not every agent needs every skill. Assign only the skills each agent's purpose requires. A content writing agent doesn't need your lead qualification skill.
Clean removal — Remove a skill from an agent without affecting the agent's other configuration. The skill remains available for other agents.
The Agent Improver Integration
Skills take reusability one step further with the Agent Improver. Agents analyse their own execution feedback and propose improvements to the skills they use.
Example flow:
- Your lead qualification agent runs 50 scoring workflows
- It identifies that leads from a specific industry are being misclassified
- The Agent Improver proposes an update to the qualification criteria
- You review and approve the change — one update, inherited by all agents using that skill
This turns skills from static modules into capabilities that improve with use.
FAQ
How is a skill different from a tool?
A tool is an external action — send an email, create a ticket, post a message. A skill is a workflow that may use tools internally. Your "lead qualification" skill uses email and CRM tools, but the skill itself is the knowledge of when and how to use them.
Can a skill be applied to multiple agents at once?
Yes. Save the skill once, then assign it to any agent. Updates propagate to all agents using it.
Do I need to code to build a skill?
No. The Agent Builder creates skills from a plain English description. For advanced customization, the API is available for developers, but it is not required.
Can agents propose skill improvements?
Yes. The Agent Improver analyses execution feedback and proposes skill updates. You review and approve before anything changes.