Back to skill

Security audit

Bring Photo MCP

Security checks for vulnerabilities and agentic risk

Overview

The skill’s shopping-list access and optional photo attachments fit its stated purpose, with privacy and external-package considerations.

Limit writable lists to those you intend to manage and store the password in a secret store. Explicitly authorize original-photo uploads and remove private details first: attached images go to Bring and may be visible to list members. All bundled files were inspected and scanner coverage was complete, but the separately installed npm runtime was not supplied for verification; review and pin that package before use.

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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · CHANGELOG.md (reported line 9)May include surrounding context.

md
### Fixed

- Long-running MCP sessions now recover from an explicit Bring HTTP 401 by reauthenticating with the configured account. Previously, restarting the MCP server was necessary after its access token became invalid.
- Concurrent requests and late responses from an older authentication generation share a single renewal instead of triggering a login storm. Successful renewal must retain the same account identity.
- Read requests (GET/HEAD) are retried at most once after renewal. Mutating requests are never automatically replayed, avoiding duplicate or uncertain shopping-list changes.

The photo-upload workflow lacks a privacy warning about sending local image contents to Bring!.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow permits attaching a user-supplied local photo without explicitly distinguishing permission to analyze it from permission to upload its contents to Bring!. A request to add the depicted product could consequently disclose the original image, including incidental personal information, to Bring! and potentially other members of the target list. Restrictions on local paths and conversational output reduce other risks but do not establish informed consent for this transfer.

Content

No source excerpt is available for this finding.