dsh-security-guard
Run the following command in DeepSeek Harness:
dsh plugin install ruanhaodong-tt/dsh-security-guard
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install ruanhaodong-tt/dsh-security-guard inside DeepSeek Harness to install; source code is available at https://github.com/ruanhaodong-tt/dsh-security-guard .
About this plugin
DeepSeek Harness has been hit by several high-severity vulnerabilities, including arbitrary code execution via the config loader (QVD-2026-52631) and remote code execution through HTTP Host header forgery (QVD-2026-57410). DSH Security Guard addresses these at the Cordis plugin layer with pure runtime interception — no source code modification required. It shadows ctx.loader.import to validate builtins and confine relative import paths, and injects a prepend listener on http.Server to reject forged loopback Host claims originating from non-loopback peers.
The plugin logs its activation on startup and accepts an optional trustedHosts array for non-loopback authorities that should be permitted. However, it explicitly cannot cover QVD-2026-52644 (VM sandbox exec escape) or QVD-2026-52646 (prompt-injection chain sandbox escape), because those fixes reside in VM-internal closures and the sandbox context factory — code paths inaccessible to any Cordis plugin. A source patch must be applied separately to close those two gaps.
This plugin is best suited for DSH administrators who want to reduce exposure to the identified CVEs without touching the harness core. It was produced with AI assistance under human supervision and has not been audited by the upstream project. Pair it with the official source patch and evaluate the risk profile before deploying in production.
Use Cases
- Production DSH instance exposed to known CVEs, requiring rapid hardening without source changes
- Core code modification is off-limits; intercept config-load and HTTP header forgery at the plugin layer
- Pair with the official source patch to close VM-sandbox gaps that no plugin can reach
Best For
- Architects responsible for DSH operations and security posture
- Security admins seeking plugin-layer mitigation without modifying core code
- DSH platform owners evaluating defense-in-depth strategy and rollout plans
Related Plugins
A network security red-team benchmark plugin for DeepSeek-V4.1/Flash featuring dual-layer kernel injection and zero-tool architecture for controlled model safety assessment.
A security red-team prompt plugin for DeepSeek-V4 that uses code-formal mapping and zero-trigger system prompts to enable unconditional output, intended for authorized testing and research only.
Adds an Auto approval mode on the official workspace-write sandbox, classifying semantic risks, asking when ambiguous, and denying destructive operations.
Undo/rollback system for DSH: config & plugin snapshots, one-click undo/redo/restore, message-level revert, secret masking, safe mode, and offline WebUI/GUI/CLI rescue tools for when DSH won't start.