Back to skill

Security audit

cargo-connection

Security checks for vulnerabilities and agentic risk

Overview

This skill manages Cargo connections, but its deletion advice could expand into removing dependent workflows, tools, or models without clear user approval.

Review the deletion workflow before installing. Require an impact list and your approval before deleting dependent workflows, tools, or models; prefer moving their connections instead. Use narrowly scoped credentials, protect CLI output containing connector configuration, and pin a reviewed CLI version. The external CLI and missing companion safety documentation were not verified.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Listing connectors matches the description, but the explicit connector removal command goes beyond connecting tools, inspecting integrations, or fixing authentication. Removing an integration is a distinct destructive capability rather than merely a supporting implementation detail.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

Most commands support the declared integration setup and maintenance purpose. However, connector removal is a distinct destructive capability that can disconnect an integration, rather than connect it, inspect it, or repair authentication. The description does not disclose this capability. Naming, rate-limit, cache, and slug-availability operations are supporting connector-management details.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx @cargo-ai/cli'.

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The bootstrap recommends executing an unpinned npm CLI, allowing the downloaded version to change without review. Although this is not an MCP server, the underlying supply-chain concern is valid: a compromised release could execute locally with access to integration credentials, and the optional reference to version-pinning guidance does not secure this standalone instruction.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx @cargo-ai/cli'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The recovery instruction recommends installing or executing an unpinned npm package, so troubleshooting can introduce a newly published, unreviewed CLI release. This is a supply-chain weakness rather than an MCP-specific issue, and a compromised release could access local credentials or authenticated Cargo resources.

Content

No source excerpt is available for this finding.

The recovery step instructs removal of dependent resources without warning that this may disrupt workflows or delete configurations.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The recovery step suggests removing dependent resources to get past a connector-deletion safeguard without warning about configuration loss or service disruption. Authorization to remove a connector does not necessarily authorize deleting the workflows, tools, or models that depend on it, so an agent following this advice could expand the destructive scope.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · references/troubleshooting.md (reported line 32)May include surrounding context.

md
| The connector's integration has no autocomplete at all             | Not every integration offers autocomplete; enter the value directly                                                           |
| `failedToGetIntegrationAutocomplete` reason          | Integration doesn't support autocomplete for this slug             | Verify the autocomplete slug exists in the integration's `uiSchema` — not all fields use autocomplete                         |
| Stale or outdated autocomplete results               | Results are cached (default 30 minutes)                            | Pass `--refresh` to bypass the cache                                                                                          |
| Used a freeform value instead of autocomplete result  | Field requires a specific value from the autocomplete result set   | Always check `uiSchema` for `IntegrationAutocompleteWidget` — if present, fetch and use a `value` from the autocomplete results |

## Integrations