Team Skills

Give the team leader direct access to skills without delegating to a member agent.

Pass a Skills instance to the skills parameter on Team. The team leader gets skill tools and system prompt snippets directly.

1from kern.agent import Agent
2from kern.models.openai import OpenAIResponses
3from kern.skills import Skills, LocalSkills
4from kern.team import Team
5
6implementer = Agent(
7    name="Implementer",
8    role="Write code based on review feedback",
9    model=OpenAIResponses(id="gpt-4o"),
10)
11
12team = Team(
13    name="Code Review Team",
14    model=OpenAIResponses(id="gpt-4o"),
15    members=[implementer],
16    skills=Skills(loaders=[LocalSkills("/path/to/skills")]),
17    instructions=[
18        "Use your skills to review code, then delegate implementation to the Implementer.",
19    ],
20)
21
22team.print_response("Review this function and improve it", stream=True)

The leader model receives:

Skill summaries in its system prompt (via get_system_prompt_snippet())
Skill tools: get_skill_instructions, get_skill_reference, get_skill_script

Team-Level vs. Member-Level Skills

Approach	When to use
`Team(skills=...)`	The leader needs domain expertise to coordinate (e.g., review standards, routing rules)
`Agent(skills=...)` on a member	A specialist agent needs the expertise to do its own work
Both	The leader needs context to delegate, and members need context to execute

How It Works

Skills on a team follow the same pattern as knowledge, memory, and tools:

skills.get_tools() adds skill tools to the leader's tool list
skills.get_system_prompt_snippet() injects skill metadata into the leader's system prompt
The leader discovers, loads, and uses skills on demand during the run

Complete Example

1from pathlib import Path
2
3from kern.agent import Agent
4from kern.models.openai import OpenAIResponses
5from kern.skills import Skills, LocalSkills
6from kern.team import Team
7
8skills_dir = Path(__file__).parent / "sample_skills"
9
10implementer = Agent(
11    name="Implementer",
12    role="Write code based on the review feedback",
13    model=OpenAIResponses(id="gpt-4o"),
14    instructions=[
15        "You write clean, well-tested Python code.",
16        "When given review feedback, produce an improved version of the code.",
17    ],
18)
19
20review_team = Team(
21    name="Code Review Team",
22    model=OpenAIResponses(id="gpt-4o"),
23    members=[implementer],
24    skills=Skills(loaders=[LocalSkills(str(skills_dir))]),
25    instructions=[
26        "You are a team leader with access to code review skills.",
27        "Use your skills to review code, then delegate implementation work to the Implementer.",
28    ],
29    markdown=True,
30    show_members_responses=True,
31)
32
33if __name__ == "__main__":
34    review_team.print_response(
35        "Review this Python code and suggest improvements, "
36        "then have the Implementer write the improved version:\n\n"
37        "```python\n"
38        "def calculate_total(items):\n"
39        "    total = 0\n"
40        "    for i in range(len(items)):\n"
41        "        total = total + items[i]['price'] * items[i]['quantity']\n"
42        "    return total\n"
43        "```",
44        stream=True,
45    )

Next Steps

Task	Guide
Create skills	Creating Skills
Load skills into agents	Loading Skills
Build teams	Building Teams

Team Skills

Team-Level vs. Member-Level Skills

How It Works

Complete Example

Next Steps

Developer Resources