Is @earendil-works/pi-coding-agent safe?
What @earendil-works/pi-coding-agent does
Pi is a terminal coding agent, published on npm as @earendil-works/pi-coding-agent. It reads, edits and writes files, runs bash commands and keeps sessions. The npm author field lists Mario Zechner, and the code lives in earendil-works/pi. You connect a model with /login, using a subscription or an API key. Its security guide says Pi acts with the permissions of the account that started it, does not ask for approval before every tool call, and runs extensions with those same permissions. Before installing, read that guide, which calls a container, virtual machine or sandbox usually the strongest option, and give Pi only the files and short-lived credentials the task needs.
- Node.js shell/command execution
- Node.js filesystem read
- Node.js filesystem write/delete
What to do: The C/C++ code was not checked for shell, network or file access, so this result is incomplete. Review that code before installing.
@earendil-works/pi-coding-agent is an npm package analyzed by SkillTotal's deterministic static scanner. The scan found no malicious indicators. Its C/C++ code was not checked for shell, network or file access, so this result is partial. It can: delegated authentication, filesystem read, filesystem write, mcp tools detected, network egress and shell execution — capabilities are what the code can do, not a verdict on intent. Risk score 0/100 (low).
@earendil-works/
Get notified when this component changes
Change alerts aren't live yet. Leave your email and we'll let you know when they launch. We use it only for this and delete it on request.
Automated static-analysis result. It can contain false positives and false negatives, and is not a claim about the intent of @earendil-works/pi-coding-agent's authors. Report a false positive.
Behavioral traits
How this component maps to the CSA agentic threat model. Descriptive — it never affects the risk score.
Findings (6)
The component can run operating-system commands or spawn processes.
import { spawn } from "node:child_process";const proc = spawn(command, args, {import { execFile, spawnSync } from "child_process";const result = spawnSync("git", ["--no-optional-locks", "symbolic-ref", "--quiet", "--short", "HEAD"], {import { execSync, spawnSync } from "child_process";const result = spawnSync(shell, commandFromStdin ? args : [...args, command], {const output = execSync(command, {import { spawn } from "child_process";const child = spawn(shellConfig.shell, commandFromStdin ? shellConfig.args : [...shellConfig.args, command], {import { spawn } from "child_process";const child = spawn(fdPath, args, { stdio: ["ignore", "pipe", "pipe"] });import { spawn } from "child_process";const child = spawn(rgPath, args, {import { spawnSync } from "node:child_process";const trashResult = spawnSync("trash", trashArgs, { encoding: "utf-8" });import { spawn } from "node:child_process";const child = spawn(editor, [...editorArgs, filePath], {import { spawn } from "child_process";const proc = spawn("tmux", ["show", "-gv", option], {import { spawn, spawnSync } from "node:child_process";const authResult = spawnSync("gh", ["auth", "status"], { encoding: "utf-8" });proc = spawn("gh", ["gist", "create", "--public=false", tmpFile]);import { spawn } from "node:child_process";const childProcess = spawn("node", [cliPath, ...args], {import { type ChildProcess, type ChildProcessByStdio, type SpawnOptions, type SpawnOptionsWithStdioTuple, type SpawnSyncOptionsWithStringEncoding, type SpawnSyncReturns, type StdioNull, type StdioPipe } from "node:child_process";Why it matters: Powerful and often legitimate — confirm the commands aren't built from untrusted input.
Fix: Confirm the command and its arguments are fully controlled and not derived from untrusted input; prefer execFile with an argument array.
The component reads files from disk.
cachedImageBase64 = fs.readFileSync(getBundledInteractiveAssetPath(IMAGE_FILENAME)).toString("base64");const pkg = JSON.parse(stripBom(fs.readFileSync(path.join(packageRoot, "package.json"), "utf-8")));
const body = fs.readFileSync(tmpFile);
dark: JSON.parse(stripBom(fs.readFileSync(darkPath, "utf-8"))),
light: JSON.parse(stripBom(fs.readFileSync(lightPath, "utf-8"))),
const content = fs.readFileSync(registeredTheme.sourcePath, "utf-8");
const content = fs.readFileSync(themePath, "utf-8");
const content = fs.readFileSync(themePath, "utf-8");
Why it matters: Usually legitimate, but worth confirming it can't be steered into reading sensitive files.
Fix: Confirm which files are read and that paths cannot be influenced by untrusted input to reach sensitive locations.
The component writes or deletes files on disk.
fs.writeFileSync(filePath, Buffer.from(image.bytes));
fs.rmSync(forkPath, { force: true });fs.writeFileSync(debugLogPath, debugData);
fs.rmSync(tempDir, { recursive: true, force: true });Why it matters: Usually legitimate, but worth confirming the paths can't be controlled by untrusted input.
Fix: Confirm which files are written/deleted and that paths cannot be influenced by untrusted input.
The component makes outbound network requests.
</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …`)){let eq=line.indexOf("=");eq>0&&(result[line.slice(0,eq)]=line.slice(eq+1))}return result}function extractExtensionVersions(ext,memory){let versionsFnName=`qjs_ext_${ext.name.replace(/-/g,"_")}_versions`,versionsFn=ext.instance.exports[v …var __require=(x=>typeof require<"u"?require:typeof Proxy<"u"?new Proxy(x,{get:(a,b)=>(typeof require<"u"?require:a)[b]}):x)(function(x){if(typeof require<"u")return require.apply(this,arguments);throw Error('Dynamic require of "'+x+'" is n …var CANCEL_MESSAGE="Login cancelled",TIMEOUT_MESSAGE="Device flow timed out",SLOW_DOWN_TIMEOUT_MESSAGE="Device flow timed out after one or more slow_down responses. This is often caused by clock drift in WSL or VM environments. Please sync …
</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …var CANCEL_MESSAGE="Login cancelled",TIMEOUT_MESSAGE="Device flow timed out",SLOW_DOWN_TIMEOUT_MESSAGE="Device flow timed out after one or more slow_down responses. This is often caused by clock drift in WSL or VM environments. Please sync …
const response = await fetch(new URL("/v1/bug-reports", options.gatewayUrl ?? getRadiusGatewayUrl()), {return await fetch(`${this.serverUrl}${path}`, init);const response = await fetch(`${this.baseUrl}${path}`, {void fetch(`https://pi.dev/api/report-install?version=${encodeURIComponent(version)}`, {const response = await fetch(url, {const response = await fetch(url, { headers: { "User-Agent": getPiUserAgent(VERSION) } });const response = await fetch(input, signal ? { ...init, signal } : init);Why it matters: Usually legitimate, but confirm the destinations are expected and no sensitive data leaves.
Fix: Confirm the destination hosts are expected and that no sensitive data is sent off-host.
An OAuth 2.0 / OpenID Connect delegated-authentication flow was detected (authorization-code / refresh-token / token-exchange grant, an OIDC authorize/discovery endpoint or id_token, or a delegation library). Tools authenticate with the end user's delegated, scoped credentials rather than a long-lived embedded service credential. (6 occurrence(s) shown as evidence).
</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …var __require=(x=>typeof require<"u"?require:typeof Proxy<"u"?new Proxy(x,{get:(a,b)=>(typeof require<"u"?require:a)[b]}):x)(function(x){if(typeof require<"u")return require.apply(this,arguments);throw Error('Dynamic require of "'+x+'" is n …</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …</html>`}function oauthSuccessHtml(message){return renderPage({title:"Authentication successful",heading:"Authentication successful",message})}function oauthErrorHtml(message,details){return renderPage({title:"Authentication failed",heading …var CANCEL_MESSAGE="Login cancelled",TIMEOUT_MESSAGE="Device flow timed out",SLOW_DOWN_TIMEOUT_MESSAGE="Device flow timed out after one or more slow_down responses. This is often caused by clock drift in WSL or VM environments. Please sync …
Fix: Delegated auth is a lower-blast-radius execution context than an embedded static credential. Confirm the requested scopes are minimal and that tokens are never logged or forwarded off-host.
An MCP tool surface (manifest or tool definitions) was found.
import{Bm25Ranker,CODEMODE_DOCS_PATH,CODEMODE_STORE_ENTRY_TYPE,DEFAULT_TOOL_SEARCH_LIMIT,combineUsage,createToolSearchDocument,getCodemodeCallableTools,parseCodemodeSource,renderToolSample,toCodemodeDeclaration,toCodemodeIdentifier,writeOut …pi.registerTool({{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/extensions/codemode/index.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAGH,OAAO,EAAE,iBAAiB,EAAE,MAAM,iCAAiC,CAAC;AAEpE,OAAO,EAAE,4BAA4B,EAAE,MAAM,WAAW,CAAC;AAWzD,SAAS,Q …pi.registerTool(definition);
pi.registerTool({ ...definition, exposure: "hidden" });pi.registerTool({ ...definition, exposure: "hidden" });pi.registerTool(definition);
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/extensions/mcp/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAEH,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,iBAAiB,EAAE,YAAY, …{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/extensions/mcp/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAEH,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,iBAAiB,EAAE,YAAY, …{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/extensions/mcp/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAEH,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,iBAAiB,EAAE,YAAY, …{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/extensions/mcp/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAEH,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,iBAAiB,EAAE,YAAY, …pi.registerTool({ ...createToolSearchToolDefinition({ tools: pi }), defaultActive: false });{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/extensions/tool-search/index.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAGH,OAAO,EAAE,8BAA8B,EAAE,MAAM,WAAW,CAAC;AAE3D,MAAM,UAAU,yBAAyB;IACxC,OAAO,CAAC,EAAE,EAAE,EAAE; …Why it matters: Just context — review which tools it offers and their permissions.
Fix: Review the declared MCP tools and their permissions.
How attackers abuse these capabilities
Interactive labs on the attack class behind the rules above. They show the technique, not anything found in @earendil-works/pi-coding-agent.
Maintain @earendil-works/pi-coding-agent?
The free Claude Code plugin scans each npm, PyPI or MCP package with this report's open-source engine before the install command runs. It runs on your machine and needs no account.
Check your own component
Run the same evidence-backed scan on any MCP server, agent skill, or package.
Scan your own componentHow we determine this: deterministic static analysis (regex + AST), evidence-anchored, no code execution. Methodology →