Unsafe deserialization of untrusted data

CriticalAV-RUNTIME-010Runtime safetyCore pack

What it finds

Data is loaded with a deserializer that can execute code (pickle, marshal, unsafe YAML, torch.load without weights_only, node-serialize).

Why it matters

Loading a crafted file or message with these functions runs the attacker's code inside the agent process — with all its credentials and tool access.

What an attacker could do: A malicious model file, cache entry, or message could give an attacker full code execution on the agent host.

How to fix it

Use data-only formats (JSON, yaml.safe_load, safetensors, torch.load(..., weights_only=True)) and never deserialize content from users, models, or the network with pickle-style loaders.

Pro shows the corrected code for your exact file, step by step, and re-scans to confirm the fix. See plans

Check details

Finding code: UNSAFE_DESERIALIZATION
How it detects: Looks for a specific risky pattern in the code or config.
Confidence: Definite: the risky setting is written plainly in the file.
Scans: Any text-based agent source or configuration (JS/TS/Python/JSON/YAML)
OWASP LLM Top 10: LLM05 - Supply Chain Vulnerabilities, LLM03 - Training Data Poisoning
NIST AI RMF: MG-2.2, MS-2.3
SOC 2: CC7.1, CC8.1

Check your agent for this

Paste your agent code or MCP config. The free scan runs all 124 checks in about 10 seconds — no account needed.

Scan an agent free Run it in CI or your editor