Security scan report

Every plugin in the registry gets an automated static scan for suspicious patterns — hardcoded secrets, data exfiltration, destructive commands, obfuscated code and more. Here's what the latest scan found, in plain language.

1,309
plugins scanned
712
low risk
84
need review
498
high risk
15
critical
low (54%) needs review (6%) high (38%) critical (2%)

The ecosystem is largely healthy. Most plugins (60%) have no serious static findings. 15 plugins show critical-risk characteristics (secrets, exfiltration, destructive or obfuscated patterns) and 498 more have critical-severity findings — review them before installing.

51 plugins not yet scanned.

What the scan looks for

The most common findings across all plugins, explained in plain language.

🖥️Shell / subprocess
879 hits

The plugin runs shell commands or spawns subprocesses. Common in CLI/terminal plugins, but worth knowing.

Usually legitimate for terminal tools — mapped to "needs review" rather than high risk.

Dynamic code execution
600 hits

The plugin evaluates code at runtime (eval, new Function). Powerful, but harder to audit.

Usually legitimate; mapped to "needs review" unless combined with other risk.

💥Destructive commands
18 hits

The plugin contains commands that can delete files or modify critical system areas (e.g. rm -rf, chmod 777).

High risk — understand exactly what it touches before running.

🕵️Obfuscated code
6 hits

Parts of the code are deliberately hard to read (encoded blobs, fromCharCode tricks), making review difficult.

Obfuscation is a red flag — what is it hiding?

🔑Hardcoded secrets
3 hits

The plugin contains what looks like an API key, token, or private key written directly in its source code.

Anyone with access to the repo could use this credential. Treat with caution.

Note: shell/subprocess and dynamic code execution are common in CLI & terminal plugins, so they're grouped under "high risk" rather than critical risk.

Critical-risk plugins

15 plugins with secrets, exfiltration, destructive or obfuscated patterns

PluginRisk
dsh-cyber-sec★ 1Critical
dsh-auto-approve★ 3Critical
dsh-clawrouter★ 2Critical
dsh-remote★ 6Critical
dsh-terminal-panel★ 2Critical
DeepSeek-Harness-linux-★ 1Critical
dsh-security-scan★ 1Critical
dsh-desktop★ 4Critical
DeepSeek-harness-wecom★ 4Critical
dsh-openclaw-acp★ 1Critical
dsh-image-to-path★ 1Critical
DeepSeek-harness-lark★ 0Critical
zeromd★ 32Critical
DeepSeek-harness-qqbot★ 5Critical
dsh-plugin-hub★ 2Critical

Automated heuristic scan — may produce false positives. Not a manual review, never an endorsement.

High-risk plugins

498 plugins with critical-severity rule hits (code execution / shell / install) · 84 more with warnings only

PluginRisk
dsh-desktop★ 21High
dsh-hdc-bridge★ 4High
DeepSeek-Harness-Desktop★ 2High
DSH-Desktop★ 18High
DSH-Plugin-Market★ 3High
dsh-wsl-workspace★ 2High
DSH-Plugins-Marketplace★ 28High
dsh-toy★ 26High
dsh-side-panel★ 17High
dsh-sentinel★ 6High
dsh-plugin-workshop★ 14High
dsh-desktop-tools★ 2High
plugin-registry★ 35High
dsh-vscode★ 2High
dsh-server-deployment★ 2High
Dcode★ 1High
superpowers-dsh★ 16High
dsh-ssh★ 4High
dsh-doctor★ 3High
dsh-plugin-market★ 2High

Showing the 20 most severe high-risk plugins (sorted by rule hits). Browse all 498 high-risk plugins — or view this week's full report.

Note: "critical" in the Rule hits column is the severity of the matched rules (dynamic code execution, shell / subprocess — common in CLI & terminal plugins). It is not the plugin's risk tier: code-exec / shell / install hits rate High, while only secrets, exfiltration, destructive commands and obfuscation rate Critical.

📖
How scanning works

What we scan, risk levels, and the four dimensions.

🔍
Browse all plugins

Every plugin's detail page shows its own scan report.

⚙️
Deep audit (planned)

Dependency CVEs, sandboxed runtime tests, container installs — coming later.

Scans are automated heuristics — they are not a manual review and never an endorsement. Install third-party plugins at your own risk.