---
id: "CVE-2026-56838"
kind: "vulnerability"
aliases: ["CVE-2026-56838"]
title: "PraisonAI: streaming recipe path skips dangerous-tool check, runs execute_command"
severity: "high"
cvss: 7.8
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-agents", "ai-tooling", "praisonai", "python"]
not_affected_if: ["patched-version", "recipe-run-only", "trusted-recipes-only"]
published_at: "2026-10-23T12:18:15+00:00"
first_seen_at: "2026-10-23T21:57:33+00:00"
updated_at: "2026-10-25T03:08:11+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-56838", "first_seen_at": "2026-10-23T21:57:33+00:00"}]
---

# PraisonAI: streaming recipe path skips dangerous-tool check, runs execute_command

PraisonAI's recipe.run_stream() path skips the dangerous-tool policy check that recipe.run() enforces, so a recipe declaring execute_command in TEMPLATE.yaml runs even when allow_dangerous_tools is False. An attacker who can get a recipe run through the streaming path can execute commands the default policy was meant to block.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- Run `pip show praisonai` and check whether the version is from `4.5.87` up to but not including `4.6.59`.
- Search your recipes and calling code for `run_stream(` to see whether untrusted or externally supplied recipes reach the streaming path.

## Fix

- Upgrade PraisonAI to `4.6.59` or later.
- Until upgraded, avoid calling `recipe.run_stream()` with recipes you do not fully trust.

## Timeline

- 2026-10-23T21:57:33+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-56838

## From the sources

PraisonAI is a multi-agent teams system. From version 4.5.87 to before version 4.6.59, PraisonAI recipe execution blocks default-denied dangerous tools unless the caller explicitly passes allow_dangerous_tools=True. The normal recipe.run() path enforces this with _check_tool_policy(). The streaming path, recipe.run_stream(), loads the same recipe, checks dependencies, and then calls _execute_recipe() without running the dangerous-tool policy check. As a result, a recipe that honestly declares execute_command in TEMPLATE.yaml requires.tools is denied by recipe.run(), but reaches the execution engine through recipe.run_stream() with allow_dangerous_tools=False. This issue has been patched in version 4.6.59.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_