Install verification: Ecosystem plugin: installs via dsh plugin add but declares no dsh.bundle and mounts no plugin capability — the L1-L5 install ladder does not rate it; the security scan still runs.Install verification details →
What it doesAI
One-click installer for optional DeepSeek Harness components: remote execution, GitHub, web research, git publishing, bug KB, Blender, compression.
- Pick components and install them in one click; credentials entered at install
- Remote execution with pre-run auto review plus a self-healing launcher
- Adds GitHub, web research, git publishing, bug KB, Blender, compression
Capability tags are AI-inferred or rule-extracted from the README (see badge) — vocabulary-limited, not install-verified. AI = inferred; README = literal keyword from the readme.
Installation
dsh plugin --profile web add github:nmhwsygxb/dsh-tools#v1.0.0Install method: GitHub · runtime test inconclusive (L5 unknown)
Verification & Compatibility○ Ecosystem plugin
Grouped by plugin version: each card lists the dsh versions actually tested against that artifact version and their install verdicts (L5 wins). Untested combinations are omitted — absence means untested, never assumed compatible. L1–L3 are a plugin-level static archive (independent of the dsh version — the same verdict on every dsh row): they sit in each row's level grid next to L4/L5, marked "static", while their evidence is stored once per plugin inside that row's "checks & evidence" instead of being duplicated per version.
External credentials (plugin-level): none detected
package manifest not collected — the declared range is unknown.
| plugin \ dsh | dsh 0.2.1-rc.2 current | dsh 0.2.1-alpha.2 |
|---|---|---|
| v1.0.0 | ○ | ○ |
v1.0.0github○ 2 ecosystem plugin2026-10-18
https://github.com/nmhwsygxb/dsh-tools
○dsh 0.2.1-rc.2current2026-10-18Ecosystem plugin
- • Node.js: v24.14.0
- • DSH: 0.2.1-rc.2
- • Plugin artifact: github · https://github.com/nmhwsygxb/dsh-tools#v1.0.0
ID: dsh-tools-5@v1.0.0@dsh0.2.1-rc.2 · profile: l3-fullscan
Issued: 2026-10-18 · expires: 2026-10-25
View checks & evidence
not present in web-profile loader inventory
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
○dsh 0.2.1-alpha.22026-10-14Ecosystem plugin
- • Node.js: v24.14.0
- • DSH: 0.2.1-alpha.2
- • Plugin artifact: github · https://github.com/nmhwsygxb/dsh-tools#v1.0.0
ID: dsh-tools-5@v1.0.0@dsh0.2.1-alpha.2 · profile: l4-sandbox
Issued: 2026-10-14 · expires: 2026-10-21
View checks & evidence
not present in web-profile loader inventory
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
Author badges
Embed the official badges in your README to showcase security & install-test status:
`[](https://www.dsh.so/artifact/dsh-tools-5/)`Security Report
Two automated layers, neither a manual review: a static scan (per-plugin rule checks) and a model-driven dynamic security audit that covers only the subset static-flagged at warn or above and never executes plugin code.
Static low, dynamic high — stricter wins → overall high.
—v1.0.0git2026-10-21v1.0.0same versionclean · gitVet is an AST static scan (npm artifact or git source).
show static findings (7)
- warningshellExec · shell
tools/01-remote-exec/remote-agent-server.js:93const out = execSync(command, { timeout: 60000, encoding: 'utf8', maxBuffer: 10 * 1024 * 1024, windowsHide: true }) - warningnetIp · network
tools/01-remote-exec/remote-agent-server.js:200console.log(`Remote Agent running on http://0.0.0.0:${PORT}`) - warningcurlDownload · network
tools/03-github/github-manager.js:186Invoke-WebRequest -Uri $env:GH_URL -OutFile $dest -UseBasicParsing -MaximumRedirection 10 - warningcurlDownload · network
tools/04-web-research/web-research.js:177Invoke-WebRequest -Uri $env:WR_URL -OutFile $dest -UseBasicParsing -MaximumRedirection 10 -Headers @{ 'User-Agent' = $env:WR_UA } - warningeval · code-exec
tools/08-blender/blender.js:993result = eval(compile(expr, '<dsh-blender-result>', 'eval'), ns) - warningshellExec · shell
tools/08-blender/blender.js:985exec(compiled, ns) - warningshellExec · shell
tools/08-blender/blender.js:1859'把任务拆成多次调用,或把长代码放进 .py 文件再用 exec(open(...).read())。');
show 4 audit findings
- highStandalone HTTP server executes arbitrary OS commands gated only by a static token over plaintext HTTP
tools/01-remote-exec/remote-agent-server.js:112When the operator runs this script it opens a server on 0.0.0.0 (`server.listen(PORT, '0.0.0.0', ...)`) and GET/POST /exec runs any OS command passed in `command`. The only control is a shared static token compared with `===`, accepted from the query string (`?token=`) or an X-Token header, over unencrypted HTTP, with `Access-Control-Allow-Origin: *`. Token leakage (plaintext transport, URLs in logs/history/proxies, wide CORS) turns this into unauthenticated remote code execution. /exec also accepts GET so a browser can be induced to call it.
- mediumBlender“restricted” mode omits `os` and relies on a bypassable regex guard
tools/08-blender/blender.js:56blender_run_python execs model-supplied Python in the host Blender process. The advertised restrictedMode only blocks a fixed module list that omits `os`, and the `guard` is a set of regexes that trivial obfuscation (string concat + getattr) defeats. Prompt-injected content can therefore still call os.system / read arbitrary files, so the hardening does not meaningfully contain it (the code itself admits 'not a sandbox').
- lowUnauthenticated /info and /ping leak host, IPs and cwd behind wildcard CORS
tools/01-remote-exec/remote-agent-server.js/info and /ping have no auth and responses carry `Access-Control-Allow-Origin: *`, so any web page the user visits can read the agent host's hostname, OS, Node version, LAN IPs and working directory, aiding targeted follow-up (e.g. knowing tokenConfigured/allowDir and reaching the LAN address).
- lowwr_fetch/wr_download fetch arbitrary http(s) URLs with no host allowlist (SSRF)
tools/04-web-research/web-research.jsThe only check is the http(s) scheme; the tool will request URLs pointing at loopback, RFC1918 hosts or cloud metadata (e.g. 169.254.169.254). If the model is steered by prompt-injected page/metadata content it can be used to probe or exfiltrate internal services reachable from the host.
Remote-agent HTTP server exposes networked arbitrary command execution behind only a static token; Blender and web-research guardrails are weaker than advertised.
Rule: the dynamic audit can only raise risk, never lower it (raise-only, never an exemption); a wrong raise is withdrawn by human review and the tier falls back to the static verdict. Overall verdict = the stricter of the two layers.
Disclaimer: the static scan and the dynamic security audit are both automated analysis, not a security guarantee. Always review what you install.
Activity
Last commit 2026-09-22 · activity: Active
• Repo created: 2026-09-13
• Stars: ★ 0 · Forks: 0
• Health: Active — committed within last 30 days
Source
GitHub: github.com/nmhwsygxb/dsh-tools
Author README
View on GitHub ↗The author's original README from the repository — fetched live, unedited (the English page loads the default README).
