Specialized Agents
Specialised agents are designed with a narrow focus and deep capability in their domain. Their system prompts, tools, and context are all optimised for one specific type of work.
6 min•By Priygop Team•Updated 2026
Examples of Specialised Agents
- Web Research Agent: specialises in finding current information using search tools and summarisation
- Code Generation Agent: writes, reviews, and debugs code with access to a code interpreter
- Data Analysis Agent: processes structured data, runs calculations, and produces insights
- Content Writing Agent: produces well-structured written content in the correct format and tone
- Email Drafting Agent: writes professional emails with the right tone for the recipient
- Fact-Checking Agent: verifies claims against trusted sources and flags inaccuracies
Specialised Agent System Prompt
Specialised Agent System Prompt
# System prompts define an agent's specialisation
research_agent_system_prompt = """
You are a Research Agent specialised in web research and information synthesis.
Your role:
- Search the web for specific, factual information
- Extract relevant data from search results
- Return concise, structured research summaries
Your constraints:
- Only report information you find directly — do not invent facts
- If you cannot find reliable information, say so explicitly
- Do not write long prose — use bullet points and structured data
- Focus only on what was asked — do not add unsolicited information
Your tools: web_search, extract_data, summarise
Output format:
{
"topic": "the topic researched",
"key_findings": ["finding 1", "finding 2"],
"sources": ["url1", "url2"],
"confidence": "high | medium | low",
"gaps": ["what could not be found"]
}
"""
reviewer_agent_system_prompt = """
You are a Review Agent specialised in quality assurance for research output.
Your role:
- Evaluate research summaries for accuracy, completeness, and clarity
- Verify key claims against your knowledge or available sources
- Score quality from 0-10 and explain your score
- Approve output (score >= 7) or request revision (score < 7)
"""
print("Research agent prompt defined:", len(research_agent_system_prompt), "chars")
print("Reviewer agent prompt defined:", len(reviewer_agent_system_prompt), "chars")Key Takeaways
- Specialised agents are designed with a narrow focus and deep capability in their domain.
- Web Research Agent: specialises in finding current information using search tools and summarisation
- Code Generation Agent: writes, reviews, and debugs code with access to a code interpreter
- Data Analysis Agent: processes structured data, runs calculations, and produces insights