PydanticAI Pre-Flight Check

PydanticAI Agent Readiness

The pre-flight check for PydanticAI applications. Detects agent loops, system prompt issues, and tool validation gaps.

Common PydanticAI Logic Flaws

Patterns that static analysis tools like linters don't catch.

Agent Loops

Agent.run() without iteration bounds in retry scenarios

System Prompt Injection

Dynamic system prompts built from user input

Tool Security

@tool decorated functions without input validation

Detection Patterns

PydanticAI-specific detection patterns with code examples.

Dynamic System Prompt Injection

CRITICAL

System prompt built from user-controlled data.

Vulnerable
python
from pydantic_ai import Agent

# System prompt includes user data - injectable!
agent = Agent(
    "openai:gpt-4",
    system_prompt=f"You help with {user_topic}. "
                  f"The user prefers {user_preference}."
)
result = await agent.run(user_query)
Secure
python
from pydantic_ai import Agent

# Static system prompt with structured deps
agent = Agent(
    "openai:gpt-4",
    system_prompt="You are a helpful assistant.",
    deps_type=UserContext,
)

@agent.system_prompt
def add_context(ctx: RunContext[UserContext]) -> str:
    # Context injected through typed deps, not strings
    return f"User topic: {ctx.deps.topic}"

Getting Started

Run Inkog against your PydanticAI codebase.

1

Run the scanner

bash
npx -y @inkog-io/cli scan ./my-pydantic-ai-app
2

Review findings

Inkog traces data flow through your PydanticAI code and reports issues with severity levels and line numbers.

3

Address issues

Apply the suggested fixes based on severity and re-scan to verify.

PydanticAI Compliance Reports

Automated mapping to global AI governance frameworks.

EU AI Act

Article 14, 15, 12

NIST AI RMF

MAP/MEASURE/MANAGE

OWASP LLM

Top 10 Coverage

ISO 42001

AI Management

PydanticAI Readiness FAQ

Does Inkog support PydanticAI?

Yes. Inkog has a Python adapter that understands PydanticAI Agent patterns, @tool decorators, system prompt construction, and structured output validation.

How does PydanticAI compare to LangChain for security?

PydanticAI uses type-safe patterns that prevent some vulnerability classes, but agent loops and prompt injection are still possible. Inkog detects these regardless of framework.

Scan Your PydanticAI Application

Free tier available. No credit card required.