Back to skill

Security audit

Pipeworx could-have-been-email

Security checks for vulnerabilities and agentic risk

Overview

This skill is a small, disclosed MCP wrapper that sends user-provided meeting transcripts to an external analysis service, so the main risk is choosing what content to submit.

Install only if you are comfortable sending meeting transcript text to the disclosed external MCP service. Avoid submitting confidential, regulated, customer, credential, or highly sensitive business content unless you have reviewed the provider's privacy and retention practices and have approval to use it.

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
Findings (1)

other

Warning
Location
SKILL.md:5
Finding

Potential Disclosure of Sensitive Meeting Transcripts to an External MCP Service

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 5–14
Vulnerability Type: External transmission of potentially sensitive data
Risk Level: Medium

Vulnerable Code Snippet

markdown
## could_have_been_email_analyze

Check if a meeting transcript could have been an email instead. Returns filler word count, decisions

```json
{
  "mcpServers": {
    "could-have-been-email": {
      "url": "https://gateway.pipeworx.io/could-have-been-email/mcp"
    }
  }
}
text

### Technical Analysis

The documented skill sends meeting-transcript content to an externally hosted MCP endpoint at `gateway.pipeworx.io` for analysis. Meeting transcripts may contain confidential business discussions, personal information, customer data, credentials, or other regulated material.

This remote processing is consistent with the skill's advertised purpose, and the reviewed file does not establish malicious intent. However, the documentation does not require informed consent, local redaction, data minimization, endpoint trust verification, or review of the provider's retention and privacy terms. It also states that an `X-API-Key` is required without documenting a secure credential-provisioning mechanism.

### Attack Path

1. A user adds the documented external MCP endpoint to the agent configuration.
2. The user or agent invokes `could_have_been_email_analyze` with a meeting transcript.
3. The transcript is transmitted outside the user's environment to `gateway.pipeworx.io`.
4. The external service receives and processes potentially sensitive content.
5. If the provider, endpoint, transport termination environment, or data-handling practices are untrusted or compromised, the transcript may be retained, inspected, or disclosed.

### Impact Assessment

The issue does not grant demonstrated local code execution, elevated system privileges, or persistence. Its scope is the confidentiality of data submitted to the
...[truncated 312 chars]
Remediation
View remediation

Remediation Suggestions

  • Clearly disclose that transcript content is transmitted to and processed by an external service.
  • Require informed user confirmation before submitting each transcript.
  • Redact credentials, personal information, customer data, and confidential business details locally before transmission.
  • Minimize submissions to only the text necessary for the requested analysis.
  • Document the provider's privacy, retention, deletion, data-location, and subprocesser policies.
  • Verify and allowlist the expected HTTPS endpoint and evaluate the provider before deployment.
  • Provision the required X-API-Key through a secret manager or protected environment variable rather than embedding it in skill files, prompts, logs, or source control.
  • Prevent API keys and transcript contents from appearing in diagnostic logs.
  • Where confidentiality requirements prohibit third-party processing, provide a locally hosted analysis option.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep