CVE-2026-56692: NanoClaw Symlink-Following Vulnerability Allows Host File Disclosure
NanoClaw versions before 2.1.17 contain a symlink-following vulnerability that allows malicious container agents to read sensitive files from the host system. When the host processes file attachments, it validates filenames using a safety check but then copies files without verifying whether those filenames point to symlinks. An attacker controlling a container agent can craft a symlink with a seemingly benign name to trick the host into copying and exposing arbitrary files it can access, leading to information disclosure.
Source data · NVD / CISA · public domain
- CVSS
- 3.1 · 5.5 MEDIUM · CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N
- Weaknesses (CWE)
- CWE-59
- Affected products
- 0 configuration(s)
- Published / Modified
- 2026-06-23 / 2026-06-24
NVD description (verbatim)
NanoClaw before 2.1.17 contains a symlink following vulnerability in forwardAttachedFiles that allows container-controlled agents to exfiltrate host-readable files. The host validates attachment filenames using only isSafeAttachmentName before copying with fs.copyFileSync, which follows symlinks without containment checks, allowing malicious agents to disclose arbitrary host files.
3 reference(s) · View on NVD →
SEC.co analysis · AI-assisted, reviewed against source
Technical summary
The vulnerability exists in NanoClaw's forwardAttachedFiles function, which validates attachment filenames through the isSafeAttachmentName function before executing fs.copyFileSync. This validation mechanism does not account for symlinks; it checks only the filename itself, not whether the path target is a symbolic link. When fs.copyFileSync follows the symlink without additional containment checks, a container-controlled agent can specify a symlink whose name passes validation but whose target points to sensitive host files, enabling arbitrary file read access. The flaw is classified as CWE-59 (Improper Link Resolution Before File Access).
Business impact
An attacker with control over a container agent can exfiltrate sensitive host-readable files, including configuration files, application secrets, credentials, or other confidential data. This information disclosure can facilitate secondary attacks, privilege escalation, or lateral movement within the infrastructure. Organizations using NanoClaw to orchestrate containerized workloads face exposure of data that should remain isolated between container and host boundaries.
Affected systems
NanoClaw versions prior to 2.1.17 are affected. The vulnerability requires the attacker to have the ability to control or influence a container agent's behavior, making it relevant primarily in multi-tenant or shared container environments where agents may be compromised or controlled by untrusted actors.
Exploitability
Exploitation requires local access (the vector shows AV:L) and low-privilege container agent control (PR:L). No user interaction is necessary (UI:N). An attacker with agency over a container agent can reliably craft a malicious attachment filename as a symlink to exploit this flaw. The attack is straightforward once the agent is compromised, but the initial compromise of the container agent itself is a prerequisite.
Remediation
Upgrade NanoClaw to version 2.1.17 or later. Organizations unable to upgrade immediately should review and restrict which container agents have permission to submit file attachments, and monitor host file system access patterns for suspicious symlink traversal attempts.
Patch guidance
Apply the upgrade to NanoClaw 2.1.17 or later as soon as possible. Verify the patch by checking the installed version and confirming that the forwardAttachedFiles function now includes symlink validation or uses safer file handling primitives (such as realpath resolution with containment checks). Test in a staging environment before production deployment to ensure compatibility with existing container orchestration workflows.
Detection guidance
Monitor for unusual symlink creation patterns within container agent working directories. Log and alert on fs.copyFileSync operations that involve symbolic links, especially when those symlinks resolve to paths outside the expected attachment directory. Examine host access logs for file reads from sensitive locations (etc, opt, home directories) that correlate with container agent activity. Network and process monitoring can flag exfiltration of file contents immediately after forwardAttachedFiles processing.
Why prioritize this
Although the CVSS score is moderate (5.5), the vulnerability represents a direct boundary violation in containerized environments. Information disclosure of host files can undermine the security model of container isolation and enable follow-on attacks. Prioritize patching in environments where untrusted or semi-trusted agents run, and defer in strictly controlled, single-tenant deployments with no agent compromise risk.
Risk score, explained
The CVSS 3.1 score of 5.5 (MEDIUM) reflects high confidentiality impact (C:H) but no integrity or availability impact. Local access and low privilege are required, limiting the immediate threat scope. However, the ability to read arbitrary host files without authentication or user interaction elevates the concern for data protection and compliance.
Frequently asked questions
Can an attacker exploit this without first compromising a container agent?
No. The vulnerability requires the ability to control or manipulate the behavior of a container agent to craft a malicious attachment filename. Direct network or host-level access to NanoClaw itself is not sufficient; the attack originates from within the container layer.
Does the isSafeAttachmentName function prevent symlink attacks?
No. The function validates the filename string itself (e.g., checking for path traversal sequences) but does not follow the link or verify that the target is a regular file rather than a symlink. This is a classic example of a time-of-check-time-of-use (TOCTOU) flaw combined with unsafe symlink handling.
What types of files are at risk?
Any file that the host process running NanoClaw can read is potentially at risk. This commonly includes configuration files, environment variable dumps, private keys, database credentials, and application secrets stored on the host filesystem.
Is there a workaround if we cannot upgrade immediately?
Implement strict access controls on container agents to limit which entities can submit file attachments. Additionally, run NanoClaw with the minimal set of host file read permissions necessary for its operation, and isolate sensitive host files to separate directories with restricted permissions. Monitor file access logs closely.
This analysis is based on the CVE record and vendor advisories current as of the publication date. Always verify patch availability and compatibility with your specific NanoClaw deployment before applying updates. The CVSS score and severity reflect the vulnerability in isolation; actual risk depends on your environment, threat model, and the trustworthiness of container agents in your deployment. For definitive remediation steps, consult the official NanoClaw security advisory and release notes for version 2.1.17 or later. Source: NVD (public-domain), retrieved 2026-07-29. Analysis generated by SEC.co (claude-haiku-4-5).
Weaknesses (CWE)
Related vulnerabilities
- CVE-2025-46293MEDIUMmacOS Symlink Traversal Allows Unauthorized Data Access
- CVE-2026-11322MEDIUMHermes WebUI Path Traversal Vulnerability – Credential Exposure Risk
- CVE-2026-11853MEDIUMDebusine Arbitrary Symlink Creation via Manifest Path Traversal
- CVE-2026-28262MEDIUMDell iDRAC Tools Symlink Follow Information Tampering Vulnerability
- CVE-2026-40861MEDIUMApache Airflow Path Traversal – Log Directory Symlink and Directory Escape Vulnerability
- CVE-2026-44275MEDIUMDell/Alienware Purchased Apps Link Following Vulnerability (CVSS 6.3)
- CVE-2026-45384MEDIUMArbitrary File Overwrite in bit7z via Symlink Attack
- CVE-2026-45491MEDIUM.NET Link-Following Vulnerability – Local File Tampering Risk