Vulnerability Database

Browse Inkog’s comprehensive database of AI agent security vulnerabilities.

Each vulnerability includes detection rules, affected frameworks, and remediation guidance.

Showing 8 of 8 vulnerabilities

INKOG-001HIGH

Prompt Injection

Unsanitized user input embedded in prompts allowing attackers to override system instructions.

LangChainLangGraphCrewAI

INKOG-002CRITICAL

Tainted Eval/Exec

LLM-generated or user-controlled code executed via eval(), exec(), or equivalent functions.

LangChainAutoGenn8n

INKOG-003CRITICAL

Hardcoded Credentials

API keys, passwords, or tokens hardcoded in source files instead of using environment variables.

All

INKOG-004CRITICAL

Infinite Loop

Loop condition depends on LLM output without deterministic termination guarantee.

LangGraphCrewAIAutoGen

INKOG-005CRITICAL

SQL Injection via LLM

LLM-generated SQL queries concatenated without parameterization allowing database attacks.

LangChainLlamaIndex

INKOG-006HIGH

Output Validation Failures

LLM output used in dangerous sinks (eval, HTML, SQL, commands) without validation.

All

INKOG-007CRITICAL

Cross-Tenant Data Leakage

Data from one tenant accessible to another due to improper isolation in multi-tenant systems.

All

INKOG-008CRITICAL

Unsafe Deserialization

Untrusted data deserialized via pickle or YAML allowing arbitrary code execution.

LangChainLlamaIndex

Vulnerability Categories

Injection Attacks

Vulnerabilities where untrusted input manipulates AI behavior:

Prompt Injection: User input alters LLM instructions
Indirect Prompt Injection: Malicious content in retrieved data
Template Injection: Dynamic template construction with user data

Authorization & Access Control

Improper access controls in agent systems:

Tool Use Without Validation: Agents execute tools without input checks
Privilege Escalation: Agents perform actions beyond intended scope
Missing Authentication: Unprotected agent endpoints

Memory & State

Vulnerabilities in agent memory and state management:

Memory Poisoning: Malicious data persisted to long-term memory
State Manipulation: Session state altered by untrusted input
Context Window Attacks: Exceeding context limits to truncate instructions

Information Disclosure

Unintended exposure of sensitive information:

Chain-of-Thought Leakage: Internal reasoning exposed to users
System Prompt Extraction: Attackers extract hidden instructions
Sensitive Data in Prompts: Credentials or PII in prompt text

Code Execution

Risks from dynamic code execution:

Unrestricted Code Interpreters: Sandbox escapes in code tools
Unsafe Deserialization: Arbitrary object instantiation
Command Injection: Shell commands with user input

Severity Levels

Level	Description	Response
Critical	Immediate exploitation possible, severe impact	Fix immediately, block deployment
High	Likely exploitable with significant impact	Fix before next release
Medium	Exploitable under specific conditions	Fix in normal development cycle
Low	Limited exploitability or impact	Track and address as time permits

OWASP LLM Top 10 Mapping

Inkog’s rules map to the OWASP Top 10 for LLM Applications :

OWASP LLM	Inkog Rules
LLM01: Prompt Injection	INKOG-001, INKOG-002, INKOG-006
LLM02: Insecure Output Handling	INKOG-008, INKOG-015
LLM03: Training Data Poisoning	Out of scope (runtime analysis)
LLM04: Model Denial of Service	INKOG-020, INKOG-021
LLM05: Supply Chain	INKOG-030, INKOG-031
LLM06: Sensitive Info Disclosure	INKOG-004, INKOG-007
LLM07: Insecure Plugin Design	INKOG-002, INKOG-010
LLM08: Excessive Agency	INKOG-011, INKOG-012
LLM09: Overreliance	Advisory only
LLM10: Model Theft	Out of scope (infrastructure)

Adding Custom Rules

Define organization-specific vulnerability patterns:

.inkog.yaml


custom_rules:
  - id: "CUSTOM-001"
    title: "Internal API key in prompt"
    description: "Internal API keys should never appear in LLM prompts"
    severity: critical
    pattern:
      type: "string_in_sink"
      sink: "llm_prompt"
      match: "INTERNAL_API_KEY_\\w+"
 
  - id: "CUSTOM-002"
    title: "Unapproved LLM provider"
    description: "Only approved LLM providers should be used"
    severity: high
    pattern:
      type: "function_call"
      disallowed:
        - "openai.*"
        - "anthropic.*"
      allowed:
        - "internal_llm.*"

API Access

Query the vulnerability database programmatically:


# List all rules
curl https://api.inkog.io/v1/rules
 
# Get specific rule
curl https://api.inkog.io/v1/rules/INKOG-001
 
# Search rules
curl "https://api.inkog.io/v1/rules?framework=langchain&severity=critical"

Contributing Rules

We welcome community contributions to the vulnerability database:

Fork the rules repository
Add your rule in YAML format
Include test cases (vulnerable and secure examples)
Submit a pull request

rules/community/my-rule.yaml


id: "COMMUNITY-001"
title: "Custom framework vulnerability"
author: "your-github-username"
description: |
  Detailed description of the vulnerability.
severity: high
frameworks:
  - custom-framework
cwe: "CWE-94"
references:
  - "https://example.com/advisory"
patterns:
  - type: "taint_flow"
    source: "http_request"
    sink: "custom_framework.execute"
tests:
  vulnerable:
    - |
      from custom_framework import execute
      execute(request.data)
  secure:
    - |
      from custom_framework import execute
      execute(validate(request.data))