Install verification: runtime verified (L5) · tested on dsh 0.2.1-rc.2 · last checked Install verification details →
What it doesAI
Visual console for an AI aggregation gateway that reverse-proxies Google Antigravity and OpenAI Codex with account rotation and failover.
- Visual control panel for an AI aggregation gateway
- Reverse proxy for Google Antigravity and OpenAI Codex
- Multi-account rotation pool, 5h/week quota dashboard, failover
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 dsh-plugin-ai-gatewayInstall method: npm · runtime verified (L5)
Verification & Compatibility
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
○ The author declares no dsh version requirement — install is not blocked by version.
| plugin \ dsh | dsh 0.2.1-rc.2 current | dsh 0.2.1-alpha.4 | dsh 0.2.1-alpha.2 |
|---|---|---|---|
| 0.2.1 | ✓ | ✗ | ✗ |
0.2.1npm✓ 1 runtime-verified2 failed2026-10-19
dsh-plugin-ai-gateway
✓dsh 0.2.1-rc.2current2026-10-19L5 · runtime verified
- • Node.js: v24.14.0
- • DSH: 0.2.1-rc.2
- • Plugin artifact: npm · dsh-plugin-ai-gateway@0.2.1
ID: dsh-plugin-ai-gateway@0.2.1@dsh0.2.1-rc.2 · profile: l4-sandbox
Issued: 2026-10-19 · expires: 2026-10-26
View checks & evidence
active in loader tree (150 entries)
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
✗dsh 0.2.1-alpha.42026-10-16L5 · failed
- • Node.js: v24.14.0
- • DSH: 0.2.1-alpha.4
- • Plugin artifact: npm · dsh-plugin-ai-gateway@0.2.1
ID: dsh-plugin-ai-gateway@0.2.1@dsh0.2.1-alpha.4 · profile: l4-sandbox
Issued: 2026-10-16 · expires: 2026-10-23
View checks & evidence
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
✗dsh 0.2.1-alpha.22026-10-15L5 · failed
- • Node.js: v24.14.0
- • DSH: 0.2.1-alpha.2
- • Plugin artifact: npm · dsh-plugin-ai-gateway@0.2.1
ID: dsh-plugin-ai-gateway@0.2.1@dsh0.2.1-alpha.2 · profile: l3-fullscan
Issued: 2026-10-15 · expires: 2026-10-22
View checks & evidence
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
Only the latest 3 published versions are shown; source-snapshot (commit hash) groups, unstated groups and older versions are omitted — the full archive stays in the data layer (snapshot/unstated groups appear only when a plugin has no published version).
Author badges
Embed the official badges in your README to showcase security & install-test status:
`[](https://www.dsh.so/artifact/dsh-plugin-ai-gateway/)``[](https://www.dsh.so/artifact/dsh-plugin-ai-gateway/)`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 medium — stricter wins → overall medium.
v0.2.1dsh manifest0.2.1npm2026-10-210.2.1same versionclean · npmVet is an AST static scan (npm artifact or git source).
show static findings (11)
- criticalshellExec · shell
lib/index.js:60const out = execSync('reg query "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings"', { - criticalshellExec · shell
lib/index.js:80const portOut = execSync('powershell -NoProfile -Command "(Get-NetTCPConnection -LocalPort 7890,7891,7892,10808,10809,2080,1080 -State Liste - warningchildProcess · code-exec
lib/index.js:1516child = spawn(cmd, args, { windowsHide: true }); - warningchildProcess · code-exec
lib/index.js:1815child = spawn(config.binPath, ["-config", config.configPath], { - warningchildProcess · code-exec
lib/index.js:1869spawnSync("taskkill", ["/PID", String(pid), "/T", "/F"], { windowsHide: true, stdio: "ignore" }); - warningnetIp · network
lib/index.js:87return `http://127.0.0.1:${portOut}`; - warningnetIp · network
lib/index.js:1245const url = new URL(req.url, "http://127.0.0.1"); - warningnetIp · network
lib/index.js:1432proxyUrl: Schema.string().default("auto").description("出海代理地址(auto 自动对齐系统代理/Clash/v2rayN,可手动填写如 http://127.0.0.1:7890,填 direct 为直连/TUN)"), - warningnetIp · network
tools/gen-config.mjs:34console.log('[gen-config] 下一步 : 启动网关后访问 http://127.0.0.1:' + port + '/v1/models 验证') - warningspawnShell · shell
tools/install-into-profile.mjs:73const r = spawnSync('dsh', cmd, { stdio: 'inherit', shell: true }) - warningspawnShell · shell
tools/install-into-profile.mjs:98const dump = spawnSync('dsh', ['--profile', profile, '--dump-config'], { encoding: 'utf8', shell: true })
show 4 audit findings
- mediumBinary install trusts caller-supplied zip/checksum paths, then extracts and auto-executes the result
lib/index.jsPOST /binary/install (and /binary/download) take zipPath, checksumsPath and version from the request body. Because the expected hash is read from the caller-chosen checksums file, the SHA-256 gate is self-referential: an attacker who can POST to the panel and drop a zip+checksums pair anywhere on disk (or point at an existing file) replaces config.binPath and the plugin then spawns it (ensureRunning / watchdog / autoStart), giving code execution as the DSH user. version is merged into a filename without the /^[0-9][\w.\-]*$/ check used elsewhere, so `ASSET_NAME(version)` + join() can redirect
- mediumOpt-in standalone ZCode proxy listens on localhost with wildcard CORS and no authentication, injecting stored credentials
lib/index.jsWhen enableZCode and zcodeStandalone are enabled, startZCodeStandaloneServer exposes /v1/chat/completions, /v1/models and /status on 127.0.0.1:8325 with no authentication; if the caller sends no usable key the handler silently forwards with the user's decrypted ZCode/GLM apiKey. sendJson() answers every response with Access-Control-Allow-Origin: * and the OPTIONS preflight allows all headers, so any web page the user visits (or any other local process) can drive paid model traffic on the user's account and read the completions; /status additionally discloses the first/last characters of the ke
- lowUnauthenticated GET /provider-plan?revealKey=1 returns the gateway API key in cleartext
lib/index.jsGET routes are not subject to the controlAllowed (loopback / allowRemoteControl) gate that guards every POST route; /provider-plan with revealKey=1 returns the full data-plane key, and /plugin-config returns describe({redactSecrets:false}). With DSH bound to a non-loopback address, the key is readable by any client that can pass the same-origin check, and by any local process regardless. The key alone cannot reach a loopback-only gateway remotely, so the practical impact is limited to key disclosure / config fingerprinting.
- lowContext-meter injection builds DOM via innerHTML from provider/JWT-derived strings
lib/client.jssetupContextMeterQuotaInjection writes quota data into innerHTML template literals. targetAcc.tier comes from the local credential file's id_token claim (chatgpt_plan_type, upper-cased) and subscriptionExpiry/countdown strings are rendered too; the other panel components correctly use React text nodes. An account file whose JWT claims carry markup (e.g. an <img onerror=...> payload) is injected into the DSH page when the user opens the context meter, i.e. DOM XSS inside the DSH origin. Requires the user to have logged in with a hostile/compromised account file, hence low.
No backdoor or exfil endpoint; real risk is an install path trusting caller-supplied zip/checksum paths plus an opt-in ZCode proxy with wildcard CORS and no auth.
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 2026-09-20 · activity: Stable
• Repo created: 2026-09-13
• Stars: ★ 0 · Forks: 0
• Health: Stable — committed within 90 days
Source
Author README
View on GitHub ↗The author's original README from the repository — fetched live, unedited (the English page loads the default README).
