Back to skill

Security audit

Swarm Signal Reader

Security checks for vulnerabilities and agentic risk

Overview

The skill appears intended to read swarm intelligence, but its setup under-discloses that it installs a mutable external MCP server with interactive posting tools.

Review this carefully before installing. The skill may be useful for reading public consensus signals, but the documented setup runs an external, unpinned MCP package and exposes tools that can post or react externally despite the read-only label. Install only if you trust the publisher and package source, and prefer a pinned version or sandboxed MCP environment with clear approval before any posting, reacting, tipping, or sensitive-query use.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:13
Finding
Unpinned Third-Party npm Package Is Downloaded and Executed Automatically## Vulnerability Details **File Location**: `SKILL.md`, lines 13-20 **Vulnerability Type**: `T08: Insecure Dependencies` **Risk Level**: Medium **Vulnerable Code Snippet**: ```markdown Add to your `.mcp.json` — no wallet, no tokens needed: ```json { "mcpServers": { "supercolony": { "command": "npx", "args": ["-y", "supercolony-mcp"] } } } ``` ``` ### Technical Analysis The documented MCP configuration invokes `npx -y supercolony-mcp` without an exact package version or integrity constraint. The `-y` option suppresses the package-installation confirmation, while the unversioned package name allows the npm registry to resolve a package release that may change after this skill has been reviewed. The project does not include the package implementation, a lockfile, a cryptographic integrity hash, or a source repository reference that would allow the remotely obtained executable code to be verified as part of this audit. Consequently, the behavior ultimately executed by users is mutable and falls outside the reviewed artifact. This is a supply-chain risk rather than evidence that the current npm package is malicious. Exploitation would require compromise or malicious control of a relevant supply-chain component, such as the package, a package release, a maintainer account, or the registry resolution path. ### Attack Path 1. A user follows the setup instructions and adds the supplied server definition to `.mcp.json`. 2. The Agent or MCP client launches the configured command. 3. `npx` resolves `supercolony-mcp` from the configured npm registry without enforcing an audited version. 4. The `-y` option permits installation without an interactive confirmation. 5. If the resolved package or supply chain has been compromised, attacker-controlled package lifecycle or runtime code executes locally. 6. That code operates with the permissions of the user running the Agent and may interact with resources available ...[truncated 718 chars]
Remediation
## Remediation Suggestions 1. Pin `supercolony-mcp` to a reviewed, exact version rather than relying on the latest registry resolution. 2. Use a lockfile and verify npm integrity metadata or a separately published cryptographic checksum before execution. 3. Document the authoritative package registry and official source-code repository so users can validate package provenance. 4. Avoid automatic installation through `npx -y`; prefer a deliberate installation and review step before configuring the MCP server. 5. Consider vendoring or otherwise distributing the reviewed implementation with reproducible build and release provenance. 6. Run the MCP server with least privilege in a sandbox or container, restrict filesystem and network access, and avoid exposing unnecessary secrets through environment variables. 7. Keep dependency versions under controlled review and establish a process for verifying updates before changing the pinned version.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

MCP Config Access

High
Category
Agent Snooping
Content
## Zero-Config Setup (Read-Only)

Add to your `.mcp.json` — no wallet, no tokens needed:

```json
{
Confidence
94% confidence
Finding
Instructing users to add a new MCP server to `.mcp.json` expands the agent's available tool surface and grants an external package runtime access within the agent environment. Because the package is fetched via `npx` and the server exposes networked capabilities, this creates supply-chain and data-boundary risk, especially when the documentation understates the write-capable nature of the tools.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The manifest description says to use the skill when you 'need market consensus, want to know what other agents think about a topic, or need swarm-validated intelligence before making a decision.' These triggers are very broad and lack clear boundaries or exclusion conditions, making unintended invocation more likely in normal decision-support conversations.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The phrase 'when you need collective intelligence' is subjective and could apply to a wide range of common user requests. The file does not provide a bounded list of trigger phrases, scope limits, or negative examples to distinguish intended from unintended activation.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
Labeling the setup as 'Zero-Config Setup (Read-Only)' while advertising interactive tools like `hive_ask`, `hive_react`, and `hive_tip` is materially misleading. This can cause operators to enable an MCP server believing it only performs passive reads, when in fact it exposes tools that can transmit data or trigger external actions.

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
The skill is presented as a read-only consensus reader, but it also instructs use of `hive_ask`, which posts a QUESTION to an external hive. That creates a capability mismatch: an agent or user may disclose prompts, strategy, or sensitive context to a third-party network under the assumption they are only reading public signals.

Static analysis

No suspicious patterns detected.