| 1 | import { mkdirSync, writeFileSync } from "node:fs"; |
| 2 | import { createServer, request, type Server } from "node:http"; |
| 3 | import net from "node:net"; |
| 4 | import path from "node:path"; |
| 5 | import { afterAll, beforeAll, describe, expect, it } from "vitest"; |
| 6 | import { bwrapArgs, JAIL_HOSTS, jailAllows, MODEL_URL, parseJailArgs, REGISTRY_HOSTS, SHARED_ENV, startJailProxy, startModelProxy } from "../src/jail.js"; |
| 7 | import { tmp } from "./helpers.js"; |
| 8 | |
| 9 | describe("a terminal task's jail", () => { |
| 10 | it("takes only task directories beside the checkout", () => { |
| 11 | const a = parseJailArgs(["--root", "/w", "agent", "--dir", "../task-1-1", "--out", "../task-1-1.out", "--cli", "claude-code", "--", "claude", "-p", "x"], "/w/repo"); |
| 12 | expect(a).toEqual({ mode: "agent", root: "/w", dir: "/w/task-1-1", out: "/w/task-1-1.out", cli: "claude-code", egress: "cli", cmd: ["claude", "-p", "x"] }); |
| 13 | const bad = (dir: string, out = "../o") => () => parseJailArgs(["--root", "/w", "agent", "--dir", dir, "--out", out, "--", "true"], "/w/repo"); |
| 14 | expect(bad(".")).toThrow(/isn't a task directory/); // the person's checkout |
| 15 | expect(bad("../repo")).toThrow(/isn't a task directory/); |
| 16 | expect(bad("../.jail")).toThrow(/isn't a task directory/); |
| 17 | expect(bad("/etc")).toThrow(/isn't a task directory/); |
| 18 | expect(bad("../a/b")).toThrow(/isn't a task directory/); |
| 19 | expect(bad("../a", "../a")).toThrow(/must differ/); |
| 20 | expect(() => parseJailArgs(["--root", "/w", "agent", "--dir", "../a", "--out", "../b"])).toThrow(/usage/); |
| 21 | expect(() => parseJailArgs(["--root", "/w", "agent", "--dir", "../a", "--out", "../b", "--cli", "gemini", "--", "x"], "/w/repo")).toThrow(/unknown --cli/); |
| 22 | }); |
| 23 | |
| 24 | it("reaches the providers' sign-in and API hosts and download hosts of package registries, by name", () => { |
| 25 | expect(jailAllows("claude.ai", 443)).toBe(false); |
| 26 | for (const h of ["api.anthropic.com", "chatgpt.com", "auth.openai.com", "registry.npmjs.org", "files.pythonhosted.org", "index.crates.io", "API.Anthropic.com."]) expect(jailAllows(h, 443), h).toBe(true); |
| 27 | // Upload hosts, fetch-for-you proxies, arbitrary sites, and other ports are not. |
| 28 | for (const h of ["upload.pypi.org", "crates.io", "proxy.golang.org", "api.openai.com", "example.com", "github.com", "evil.anthropic.com.example.com"]) expect(jailAllows(h, 443), h).toBe(false); |
| 29 | expect(jailAllows("api.anthropic.com", 80)).toBe(false); |
| 30 | expect(jailAllows("api.anthropic.com", 22)).toBe(false); |
| 31 | expect(jailAllows("archive.ubuntu.com", 80)).toBe(true); |
| 32 | expect(jailAllows("registry.npmjs.org", 80)).toBe(false); |
| 33 | expect(JAIL_HOSTS).not.toContain("*"); |
| 34 | }); |
| 35 | |
| 36 | it("hides the home directory, the work root and /tmp, and shows only the task's clone, its output, the checkout's objects and the CLI's sign-in", () => { |
| 37 | const a = parseJailArgs(["--root", "/w", "agent", "--dir", "../t", "--out", "../t.out", "--cli", "claude-code", "--", "claude"], "/w/repo"); |
| 38 | const tools = tmp("jail-tools-"); |
| 39 | mkdirSync(path.join(tools, "home")); |
| 40 | writeFileSync(path.join(tools, "home", ".claude_.credentials.json"), "{}"); |
| 41 | const args = bwrapArgs(a, { home: "/home/p", tools, env: { PATH: "/home/p/bin:/usr/bin", NODE_EXTRA_CA_CERTS: "/etc/ca.crt", DIFFINERY_JOB_TOKEN: "secret", GIT_AUTHOR_NAME: "bot" }, cliPath: "/usr/lib/claude/cli.js", nodePath: "/usr/bin/node" }); |
| 42 | const s = args.join(" "); |
| 43 | for (const f of ["--unshare-all", "--die-with-parent", "--new-session", "--clearenv"]) expect(args).toContain(f); |
| 44 | expect(s).toContain("--cap-drop ALL"); |
| 45 | expect(args).not.toContain("--share-net"); |
| 46 | // /run too: the Docker daemon's and BuildKit's sockets live there, and a socket file reaches across network namespaces. |
| 47 | for (const h of ["/tmp", "/run", "/home/p", "/w"]) expect(s).toContain(`--tmpfs ${h}`); |
| 48 | expect(s).toContain("--bind /w/t /w/t"); |
| 49 | expect(s).toContain("--bind /w/t.out /w/t.out"); |
| 50 | expect(s).toContain("--ro-bind /w/repo/.git/objects /w/repo/.git/objects"); |
| 51 | expect(s).not.toMatch(/\/w\/repo(\/\.git)?(\/config)? /); |
| 52 | expect(s).not.toContain("--bind /w/repo/.git "); |
| 53 | // The environment is cleared, then given only what it needs: no sandbox secrets. |
| 54 | expect(s).not.toContain("secret"); |
| 55 | expect(s).toContain("--setenv GIT_AUTHOR_NAME bot"); |
| 56 | expect(s).toContain("--setenv HTTPS_PROXY http://127.0.0.1:3128"); |
| 57 | expect(s).toContain(`--setenv PATH ${tools}/bin:/usr/bin`); |
| 58 | // The sign-in is a private copy, and the jail's tools are read-only. |
| 59 | expect(s).toContain(`--bind ${tools}/home/.claude_.credentials.json /home/p/.claude/.credentials.json`); |
| 60 | expect(s).not.toContain("--bind /home/p/.claude/.credentials.json"); |
| 61 | expect(s).toContain(`--ro-bind ${tools} ${tools}`); |
| 62 | expect(s).not.toContain(".claude.json"); // not copied here, so not bound |
| 63 | expect(s).toContain("--chdir /w/t"); |
| 64 | // A finish jail has no way out and no sign-in. |
| 65 | const f = bwrapArgs({ ...a, mode: "finish", cli: undefined }, { home: "/home/p", tools, env: {}, nodePath: "/usr/bin/node" }).join(" "); |
| 66 | expect(f).not.toContain("HTTPS_PROXY"); |
| 67 | expect(f).not.toContain(".claude"); |
| 68 | }); |
| 69 | |
| 70 | describe("its proxy", () => { |
| 71 | let dir: string; |
| 72 | let sock: string; |
| 73 | let close: () => void; |
| 74 | const logs: string[] = []; |
| 75 | beforeAll(async () => { |
| 76 | dir = tmp("jail-"); |
| 77 | sock = path.join(dir, "p.sock"); |
| 78 | const server = await startJailProxy(sock, (l) => logs.push(l)); |
| 79 | close = () => server.close(); |
| 80 | }); |
| 81 | afterAll(() => close()); |
| 82 | |
| 83 | const connect = (target: string) => |
| 84 | new Promise<string>((resolve) => { |
| 85 | const c = net.connect(sock, () => c.write(`CONNECT ${target} HTTP/1.1\r\nHost: ${target}\r\n\r\n`)); |
| 86 | let got = ""; |
| 87 | c.on("data", (d) => { |
| 88 | got += d; |
| 89 | if (got.includes("\r\n\r\n")) { |
| 90 | c.destroy(); |
| 91 | resolve(got.split("\r\n")[0]!); |
| 92 | } |
| 93 | }); |
| 94 | c.on("end", () => resolve(got.split("\r\n")[0]!)); |
| 95 | c.on("error", () => resolve(got.split("\r\n")[0] ?? "")); |
| 96 | }); |
| 97 | |
| 98 | it("refuses tunnels anywhere else, and to the allowed hosts on other ports or by address", async () => { |
| 99 | expect(await connect("example.com:443")).toBe("HTTP/1.1 403 Forbidden"); |
| 100 | expect(await connect("api.anthropic.com:80")).toBe("HTTP/1.1 403 Forbidden"); |
| 101 | expect(await connect("1.1.1.1:443")).toBe("HTTP/1.1 403 Forbidden"); |
| 102 | expect(await connect("api.anthropic.com.evil.test:443")).toBe("HTTP/1.1 403 Forbidden"); |
| 103 | expect(logs.some((l) => l.includes("refused a connection to example.com:443"))).toBe(true); |
| 104 | }); |
| 105 | |
| 106 | it("serves plain HTTP only to package mirrors, and only downloads", async () => { |
| 107 | const send = (method: string, url: string) => |
| 108 | new Promise<number>((resolve) => { |
| 109 | const r = request({ socketPath: sock, method, path: url, headers: { host: new URL(url).host } }, (res) => { |
| 110 | res.resume(); |
| 111 | resolve(res.statusCode ?? 0); |
| 112 | }); |
| 113 | r.on("error", () => resolve(0)); |
| 114 | r.end(); |
| 115 | }); |
| 116 | expect(await send("GET", "http://example.com/")).toBe(403); |
| 117 | expect(await send("POST", "http://archive.ubuntu.com/ubuntu/")).toBe(403); |
| 118 | expect(await send("GET", "http://registry.npmjs.org/left-pad")).toBe(403); |
| 119 | }); |
| 120 | }); |
| 121 | }); |
| 122 | |
| 123 | describe("a headless run's jail (phase 5d)", () => { |
| 124 | const args = (...extra: string[]) => ["--root", "/w", "agent", "--dir", "../task-2-1", "--out", "../task-2-1.out", ...extra, "--", "claude", "-p", "x"]; |
| 125 | it("takes --egress model, for agent jails without a sign-in", () => { |
| 126 | expect(parseJailArgs(args("--egress", "model"), "/w/repo").egress).toBe("model"); |
| 127 | expect(() => parseJailArgs(args("--egress", "model", "--cli", "claude-code"), "/w/repo")).toThrow(/without --cli/); |
| 128 | expect(() => parseJailArgs(["--root", "/w", "finish", "--dir", "../a", "--out", "../b", "--egress", "model", "--", "true"], "/w/repo")).toThrow(/for agent jails/); |
| 129 | expect(() => parseJailArgs(args("--egress", "web"), "/w/repo")).toThrow(/unknown --egress/); |
| 130 | }); |
| 131 | |
| 132 | it("reaches the package registries only: no CLI vendor, and never Diffinery's host", () => { |
| 133 | for (const h of ["api.anthropic.com", "console.anthropic.com", "chatgpt.com", "auth.openai.com", "diffinery.example"]) expect(jailAllows(h, 443, "model"), h).toBe(false); |
| 134 | for (const h of REGISTRY_HOSTS) expect(jailAllows(h, 443, "model"), h).toBe(true); |
| 135 | expect(jailAllows("archive.ubuntu.com", 80, "model")).toBe(true); |
| 136 | }); |
| 137 | |
| 138 | it("keeps the model variables, never the run's token, and sends model requests around the CONNECT proxy", () => { |
| 139 | const a = parseJailArgs(args("--egress", "model"), "/w/repo"); |
| 140 | const tools = tmp("jail-tools-"); |
| 141 | const env = { PATH: "/usr/bin", ANTHROPIC_BASE_URL: `${MODEL_URL}/-/model/anthropic`, ANTHROPIC_API_KEY: "diffinery-jail", DIFFINERY_MODEL_TOKEN: "dmt_repo_secret", DIFFINERY_MODEL_ORIGIN: "https://d.example", OPENAI_API_KEY: "diffinery-jail" }; |
| 142 | const s = bwrapArgs(a, { home: "/home/p", tools, env, nodePath: "/usr/bin/node" }).join(" "); |
| 143 | expect(s).toContain(`--setenv ANTHROPIC_BASE_URL ${MODEL_URL}/-/model/anthropic`); |
| 144 | expect(s).toContain("--setenv ANTHROPIC_API_KEY diffinery-jail"); |
| 145 | expect(s).toContain("--setenv OPENAI_API_KEY diffinery-jail"); |
| 146 | expect(s).not.toContain("dmt_repo_secret"); |
| 147 | expect(s).not.toContain("d.example"); |
| 148 | expect(s).toContain("--setenv NO_PROXY 127.0.0.1,localhost"); |
| 149 | // A terminal task's jail keeps none of them. |
| 150 | const t = bwrapArgs({ ...a, egress: "cli", cli: "claude-code" }, { home: "/home/p", tools, env, nodePath: "/usr/bin/node" }).join(" "); |
| 151 | expect(t).not.toContain("ANTHROPIC_BASE_URL"); |
| 152 | expect(t).not.toContain("NO_PROXY"); |
| 153 | }); |
| 154 | |
| 155 | it("shares a test pool's directories, and refuses one that would undo what the jail hides", () => { |
| 156 | const e2e = tmp("jail-e2e-"); |
| 157 | mkdirSync(path.join(e2e, "pool", "s1", "repo"), { recursive: true }); |
| 158 | mkdirSync(path.join(e2e, "work")); |
| 159 | const root = path.join(e2e, "pool", "s1"); |
| 160 | const home = path.join(e2e, "home"); |
| 161 | mkdirSync(path.join(home, "sub"), { recursive: true }); |
| 162 | const work = path.join(e2e, "work"); |
| 163 | const a = parseJailArgs(["--root", root, "agent", "--dir", "../t", "--out", "../t.out", "--egress", "model", "--", "true"], path.join(root, "repo")); |
| 164 | const args = (shared?: string) => bwrapArgs(a, { home, tools: tmp("jail-tools-"), env: shared === undefined ? {} : { [SHARED_ENV]: shared }, nodePath: "/usr/bin/node" }).join(" "); |
| 165 | const s = args(work); |
| 166 | expect(s.indexOf(`--bind ${work} ${work} `)).toBeGreaterThan(s.indexOf("--tmpfs /tmp ")); |
| 167 | expect(args()).not.toContain(`--bind ${work} `); |
| 168 | // `/`, the work root (with the checkout and its git credential), the home directory, /tmp, anything holding them |
| 169 | // or inside the work root or home: shared, they would undo the jail's isolation, so the jail doesn't start. |
| 170 | for (const bad of ["/", e2e, path.join(e2e, "pool"), root, path.join(root, "repo"), home, path.join(home, "sub"), "/tmp", `${work}:/`]) expect(() => args(bad), bad).toThrow(/is, holds or is inside a directory the jail hides/); |
| 171 | for (const bad of ["relative", "/nonexistent"]) expect(() => args(bad), bad).toThrow(/isn't an existing absolute directory/); |
| 172 | }); |
| 173 | |
| 174 | describe("its model proxy", () => { |
| 175 | let sock: string; |
| 176 | let upstream: Server; |
| 177 | let close: () => void; |
| 178 | const seen: Array<{ method: string; url: string; headers: Record<string, unknown>; body: string }> = []; |
| 179 | const logs: string[] = []; |
| 180 | let slowOpen = 0; |
| 181 | let slowClosed = 0; |
| 182 | const until = async (ok: () => boolean) => { |
| 183 | for (let i = 0; i < 100 && !ok(); i++) await new Promise((r) => setTimeout(r, 20)); |
| 184 | expect(ok()).toBe(true); |
| 185 | }; |
| 186 | beforeAll(async () => { |
| 187 | upstream = createServer((req, res) => { |
| 188 | // A slow answer: it streams until whoever asked goes away. |
| 189 | if (req.headers["content-length"] === "4") { |
| 190 | slowOpen++; |
| 191 | res.writeHead(200, { "content-type": "text/event-stream" }); |
| 192 | res.write("data: thinking\n\n"); |
| 193 | res.on("close", () => slowClosed++); |
| 194 | return; |
| 195 | } |
| 196 | let body = ""; |
| 197 | req.on("data", (d) => (body += d)); |
| 198 | req.on("end", () => { |
| 199 | seen.push({ method: req.method!, url: req.url!, headers: req.headers, body }); |
| 200 | res.writeHead(200, { "content-type": "text/event-stream", "set-cookie": "s=1" }); |
| 201 | res.write("data: one\n\n"); |
| 202 | setTimeout(() => res.end("data: two\n\n"), 20); |
| 203 | }); |
| 204 | }); |
| 205 | await new Promise<void>((r) => upstream.listen(0, "127.0.0.1", r)); |
| 206 | const origin = `http://127.0.0.1:${(upstream.address() as net.AddressInfo).port}`; |
| 207 | sock = path.join(tmp("jail-model-"), "m.sock"); |
| 208 | const server = await startModelProxy(sock, origin, "dmt_repo_runtoken", (l) => logs.push(l)); |
| 209 | close = () => { |
| 210 | server.close(); |
| 211 | upstream.close(); |
| 212 | }; |
| 213 | }); |
| 214 | afterAll(() => close()); |
| 215 | |
| 216 | const send = (method: string, url: string, headers: Record<string, string> = {}, body?: string) => |
| 217 | new Promise<{ status: number; body: string; headers: Record<string, unknown> }>((resolve) => { |
| 218 | const r = request({ socketPath: sock, method, path: url, headers: { host: "127.0.0.1:3129", ...headers } }, (res) => { |
| 219 | let b = ""; |
| 220 | res.on("data", (d) => (b += d)); |
| 221 | res.on("end", () => resolve({ status: res.statusCode ?? 0, body: b, headers: res.headers })); |
| 222 | }); |
| 223 | r.on("error", () => resolve({ status: 0, body: "", headers: {} })); |
| 224 | r.end(body); |
| 225 | }); |
| 226 | |
| 227 | it("forwards model requests with the run's token in place of whatever the agent sent, and streams the answer", async () => { |
| 228 | const r = await send("POST", "/-/model/anthropic/v1/messages?beta=true", { "x-api-key": "diffinery-jail", authorization: "Bearer planted-diffinery-token", cookie: "dfy_session=x", "content-type": "application/json", "anthropic-version": "2023-06-01" }, '{"max_tokens":5}'); |
| 229 | expect(r.status).toBe(200); |
| 230 | expect(r.body).toBe("data: one\n\ndata: two\n\n"); |
| 231 | expect(r.headers["set-cookie"]).toBeUndefined(); |
| 232 | const got = seen.at(-1)!; |
| 233 | expect(got.url).toBe("/-/model/anthropic/v1/messages?beta=true"); |
| 234 | expect(got.body).toBe('{"max_tokens":5}'); |
| 235 | expect(got.headers.authorization).toBe("Bearer dmt_repo_runtoken"); |
| 236 | expect(got.headers["x-api-key"]).toBeUndefined(); |
| 237 | expect(got.headers.cookie).toBeUndefined(); |
| 238 | expect(got.headers["anthropic-version"]).toBe("2023-06-01"); |
| 239 | }); |
| 240 | |
| 241 | it("cancels the request to Diffinery when the agent hangs up", async () => { |
| 242 | const r = request({ socketPath: sock, method: "POST", path: "/-/model/anthropic/v1/messages", headers: { host: "127.0.0.1:3129" } }); |
| 243 | r.on("error", () => {}); |
| 244 | r.end("slow"); |
| 245 | await until(() => slowOpen === 1); |
| 246 | r.destroy(); |
| 247 | await until(() => slowClosed === 1); |
| 248 | }); |
| 249 | |
| 250 | it("refuses everything else on Diffinery, however the path is spelled", async () => { |
| 251 | const before = seen.length; |
| 252 | for (const url of ["/api/v3/user", "/owner/repo.git/info/refs?service=git-receive-pack", "/-/model/../api/v3/user", "/-/model/%2e%2e/api/v3/user", "/-/model/anthropic/%2E%2E/%2E%2E/api", "/-/model%2fanthropic", "/-/model/a\\..\\..\\api", "//evil.example/-/model/x", "http://evil.example/-/model/anthropic/v1/messages", "/-/modelx/anthropic"]) { |
| 253 | expect((await send("GET", url)).status, url).toBe(403); |
| 254 | } |
| 255 | expect(seen.length).toBe(before); |
| 256 | expect(logs.some((l) => l.includes("refused GET /api/v3/user"))).toBe(true); |
| 257 | }); |
| 258 | }); |
| 259 | }); |