registry / AI-agent capability abuse / rally-project-brain-mcp / 0.1.12

rally-project-brain-mcp@0.1.12

MCP server exposing Rally Project Brain to any MCP-compatible agent (Claude Code, Cursor, Copilot agent mode, Windsurf, ...)

AI Security Review

scanned 2mo ago · by lpm-firewall-ai

This is suspicious because it deliberately configures broad AI-agent control surfaces and enables activity reporting. It is not malicious under the install policy because there is no automatic install lifecycle trigger and setup requires an explicit user command.

Static reason

One or more suspicious static signals were detected.

Trigger

The user runs rally-project-brain-mcp login and provides an agent token.

Impact

Future agent sessions may run package-installed hooks and send prompts, tool inputs, and outputs to Rally.

Mechanism

Explicit setup of MCP servers, hooks, and agent instructions.

AI rationale

Evidence

index.mjs.mcp.json.cursor/mcp.json.gemini/settings.json~/.codeium/windsurf/mcp_config.json~/.codex/config.toml.vscode/mcp.json.claude/hooks/rally-common.mjs.claude/hooks/session-start.mjs.claude/hooks/user-prompt-submit.mjs.claude/hooks/post-tool-use.mjs.claude/hooks/session-end.mjs.claude/settings.jsonCLAUDE.mdAGENTS.mdGEMINI.md.cursorrules

Network endpoints1

rally-backend.fly.dev

Decision evidence

public snapshot

AI called this Suspicious at 94.0% confidence as Dangerous Capability with low false-positive risk.

Evidence for warning

The explicit login command writes MCP configuration for multiple agent products, including global Windsurf and Codex settings.
Login installs Claude Code hook scripts and enables hooks for session start, prompt submission, tool use, and session end.
The installed hooks can collect prompts, tool inputs, and tool responses for transmission to the configured Rally backend.
The package embeds hook source as Base64, reducing reviewability.

Evidence against

package.json has no preinstall, install, postinstall, or other lifecycle script.
The mutation path runs only when the user invokes the login subcommand and supplies a token.
Network use is directed to the configured Rally backend and uses the supplied actor token.

Behavioral surface

Source

EnvironmentVarsFilesystemNetwork

Supply chain

HighEntropyStringsUrlStrings

Manifest

NoLicense

Source & flagged code

6 flagged

index.mjs#virtual:base64:round1View file

4import { readFileSync, writeFileSync, existsSync, mkdirSync, unlinkSync } from "node:fs"; L5: import { execFileSync } from "node:child_process"; L6: import { join } from "node:path"; ... L8: L9: export const BASE_URL = process.env.RALLY_BACKEND_URL || "http://localhost:8080"; L10: export const PROJECT_ID = process.env.RALLY_PROJECT_ID;

High

Same File Env Network Execution

A single source file combines environment access, network access, and code or shell execution; review context before blocking.

index.mjs#virtual:base64:round1View on unpkg · L4

index.mjsView file

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. g anything else, and tell the ` + `human which one you're using by default for reporting — don't wait to be ` + `asked which identity you are; that confusion is a known, recurring problem ` + `and this is the fix for it. starting a new task? call \`create_agent_job\` ` + `once, as your first action, ` + `with a short specific title as \`goal\` formatted as "{the operator's name, ` + `if known}|{your product name, e.g. claude code/cursor/gemini cli/codex}|{a ` + `few words on the task}" (e.g. "devansh|claude code|review rally project ` + `context") — not the human's raw message verbatim, and not one call per ` + `message: keep using the same jo

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

index.mjsView on unpkg

17L18: // Embedded Claude Code hook file contents, base64-encoded to avoid any L19: // escaping hazard from the hooks' own template-literal syntax. Generated ... L63: async function runLoginAgent() { L64: const backendUrl = process.env.RALLY_BACKEND_URL || "https://rally-backend.fly.dev"; L65: const rl = createInterface({ input: process.stdin, output: process.stdout }); L66: let token; ... L87: L88: const outPath = path.join(process.cwd(), ".mcp.json"); L89: ... L250: // Cursor/Gemini/Windsurf, just a deeper OS-specific path. Unlike the L251: // targets above, this reaches into another extension's private storage

High

Base64 Obscured Url

Source decodes a Base64-obscured HTTP endpoint at runtime.

index.mjsView on unpkg · L17

154// Cursor reads its own project-scoped MCP config (.cursor/mcp.json). await writeMcpServersJson(path.join(process.cwd(), ".cursor", "mcp.json"), ".cursor/mcp.json"); // Gemini CLI reads its own project-scoped settings.json — same // mcpServers key, alongside whatever other Gemini settings (model, // theme, ...) already live in that file. await writeMcpServersJson(path.join(process.cwd(), ".gemini", "settings.json"), ".gemini/settings.json"); // Windsurf's Cascade reads one global (not project-scoped) config, // identical mcpServers schema. await writeMcpServersJson(

Medium

Ai Review Evidence

The explicit login command writes MCP configuration for multiple agent products, including global Windsurf and Codex settings.

index.mjsView on unpkg · L154

279const hooksDir = path.join(process.cwd(), ".claude", "hooks"); await mkdir(hooksDir, { recursive: true }); for (const [name, b64] of Object.entries(CLAUDE_CODE_HOOKS_B64)) { await writeFile(path.join(hooksDir, name), Buffer.from(b64, "base64")); }

Medium

Ai Review Evidence

Login installs Claude Code hook scripts and enables hooks for session start, prompt submission, tool use, and session end.

index.mjsView on unpkg · L279

index.mjs#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 1

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

index.mjs#virtual:normalized:round1View on unpkg

Source & flagged code

6 flagged · loading source

index.mjs#virtual:base64:round1View file

High

Same File Env Network Execution

A single source file combines environment access, network access, and code or shell execution; review context before blocking.

index.mjs#virtual:base64:round1View on unpkg · L4

index.mjsView file

Findings

4 High4 Medium4 Low

HighSame File Env Network Executionindex.mjs#virtual:base64:round1

HighTrigger Reachable External Ai Agent Control Surface Mutationindex.mjs

HighBase64 Obscured Urlindex.mjs

HighSemantic Analysis Limitedindex.mjs#virtual:normalized:round1

MediumNetwork

MediumEnvironment Vars

MediumAi Review Evidenceindex.mjs

LowFilesystem

LowHigh Entropy Strings

LowUrl Strings

LowNo License

Affected versions and remediation

This report applies to rally-project-brain-mcp@0.1.12.

See all versions of rally-project-brain-mcp for other recorded verdicts.

Review the evidence and your use of rally-project-brain-mcp@0.1.12 before allowing it. Restrict the permissions described in this report.
Choose an independently verified alternative or release. This report does not establish that other versions are safe.

Evidence last updated: 2026-09-03 17:25:40.097Z (UTC).

Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.