Verevoir MCP Server

This server (without client built-ins) exposes a complete data-exfiltration chain: get_card ⇒ find_governance → refine_start. Untrusted input is ingested, private data is read, and it can be sent to an external sink — and at least one leg is a direct schema wire (⇒), where a producer's output drops straight into a free-text parameter of the next tool, so the chain needs little agent cooperation. Static analysis proves the primitive exists, not that a specific run will occur.

Fix: Remove one leg of the trifecta: isolate untrusted-input tools from secret-reading tools and from egress tools, or require human approval between them.

Location: flow get_card → find_governance → refine_start

Tool "refine_start" appears to run shell commands or evaluate code (parameter "eval"). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.

Fix: Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.

Location: tool refine_start

Tool "search_start" appears to run shell commands or evaluate code (parameter "eval"). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.

Fix: Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.

Location: tool search_start

In the server's implementation (`dist/tools/loop.js:44`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: d-eval scorer = new Function('output', `return (${expression});`); } catch (err) { const message

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/tools/loop.js

In the server's implementation (`dist/card-reconcile-bin.js:2`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ecFileSync } from 'node:child_process'; import { pickWorkflowAdapter, resolveWorkflowEnv } from './router.js'; import {

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/card-reconcile-bin.js

In the server's implementation (`dist/review-bin.js:2`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: ecFileSync } from 'node:child_process'; import { FOUNDATIONAL, provisionPractices, isClean, } from '@verevoir/recipes/en

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/review-bin.js

In the server's implementation (`dist/router.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: { execSync } from 'node:child_process'; import { envFromProcessEnv } from '@verevoir/sources'; import { envFromTrelloPro

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/router.js

Tool "write_file" can write, overwrite or delete files (keyword "write_file" in tool name). Verify it is scoped to a safe directory.

Fix: Constrain file operations to an explicit, non-sensitive root; reject path traversal.

Location: tool write_file

Tool "edit_file" can write, overwrite or delete files (keyword "edit_file" in tool name). Verify it is scoped to a safe directory.

Fix: Constrain file operations to an explicit, non-sensitive root; reject path traversal.

Location: tool edit_file

Tool "refine_start" takes a command-shaped parameter "eval" with no enum/pattern constraint. Free-form, model- or attacker-controlled arguments reaching a shell is the command-injection precondition.

Fix: Constrain the parameter (enum/pattern), or build the command from a fixed template with escaped args.

Location: tool refine_start · inputSchema.properties.eval

Tool "search_start" takes a command-shaped parameter "eval" with no enum/pattern constraint. Free-form, model- or attacker-controlled arguments reaching a shell is the command-injection precondition.

Fix: Constrain the parameter (enum/pattern), or build the command from a fixed template with escaped args.

Location: tool search_start · inputSchema.properties.eval

Tool "refine_start" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool refine_start

Tool "search_start" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool search_start

Tool "write_file" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool write_file

Tool "edit_file" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.

Fix: Declare accurate annotations, and gate destructive tools on user confirmation regardless.

Location: tool edit_file

Tool "read_file" takes a path parameter "path" with no constraint. Without a canonicalize-and-contain check (not visible statically), this permits ../ traversal outside the intended root.

Fix: Resolve and verify the path stays within an allowed root; reject traversal sequences.

Location: tool read_file · inputSchema.properties.path

Tool "write_file" takes a path parameter "path" with no constraint. Without a canonicalize-and-contain check (not visible statically), this permits ../ traversal outside the intended root.

Fix: Resolve and verify the path stays within an allowed root; reject traversal sequences.

Location: tool write_file · inputSchema.properties.path

Tool "edit_file" takes a path parameter "path" with no constraint. Without a canonicalize-and-contain check (not visible statically), this permits ../ traversal outside the intended root.

Fix: Resolve and verify the path stays within an allowed root; reject traversal sequences.

Location: tool edit_file · inputSchema.properties.path