dsh-plugin-security-audit
Run the following command in DeepSeek Harness:
dsh plugin install chendengyuanxm/dsh-plugin-security-audit
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install chendengyuanxm/dsh-plugin-security-audit (repository: https://github.com/chendengyuanxm/dsh-plugin-security-audit) in DSH (v0.1.0-rc.6+) and restart your profile to activate the plugin.
About this plugin
Dynamic Cordis plugins in DSH run as plain JavaScript with direct access to real Host services — files, shell, sessions, credentials. The sandbox guards against accidental misuse; it is not a security boundary against malicious code. dsh-plugin-security-audit layers a deterministic, zero-token static audit over every cordis_define and cordis_run call, ensuring each dynamic plugin package passes through a 16-rule engine before it is minted.
The workflow has three stages: scan on define, where source is checked rule-by-rule before the package lands; gate on activate, where packages carrying critical or high findings are escalated to user approval with the top findings as the reason; and report in-stream, where a compact risk summary is appended as an additionalContexts notice beside every tool result so both model and user see it. By design the audit layer never blocks: internal errors are logged and the audited call proceeds untouched, so the auditor can never become a failure point for the harness.
Best suited for teams or individuals who already use dynamic plugins in DSH or who plan to let an agent generate Cordis packages on the fly inside a session. If any agent in your profile can touch real Host services, this zero-overhead static audit is the lowest-cost first line of defense.
Use Cases
- Automatically scan and report risks whenever an agent dynamically defines a Cordis plugin in-session.
- Escalate activation to user approval with top findings when a package contains high-severity calls like eval or hardcoded secrets.
- Apply a unified zero-token static audit layer across multi-agent profiles without altering existing tool-pipeline workflows.
Best For
- Developers and teams using dynamic Cordis plugins in DSH.
- Engineers who let agents generate and run Host code on the fly and care about security boundaries.
- Operators who want a zero-token static guard over plugin runtimes without extra cost.
Related Plugins
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.
Packages all 87 SKILL.md files from upstream reverse-skill as a DeepSeek Harness plugin that auto-registers them for authorized reverse engineering, penetration testing, and security research.