Back to skill

Security audit

SentVia Email

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed real-email skill, but it needs Review because it can send mail and create persistent external mailbox state without strong user-control guardrails and recommends an unpinned MCP install.

Install only if you are comfortable giving the agent a real SentVia mailbox and the ability to transmit email externally. Pin and review the MCP package before use, limit the API key's permissions if SentVia supports it, and require explicit confirmation for creating inboxes, sending or replying to mail, forwarding messages, and changing sender block lists.

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:24
Finding

Unpinned Third-Party Package Execution

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 24
Vulnerability Type: Unpinned npm dependency executed through npx
Risk Level: Medium

Vulnerable Code

text
openclaw mcp add sentvia --command npx --arg -y --arg sentvia-mcp

Technical Analysis

The setup instruction configures an MCP server by downloading and executing the sentvia-mcp npm package without specifying a version or verifying its integrity. The -y option also suppresses the normal installation confirmation.

Because the package source is not included in the audited project and no version or integrity hash is specified, the code executed by this command can change after the skill has been reviewed. This creates a supply-chain trust boundary around the npm registry, the package publisher account, and the package's transitive dependencies.

This is an insecure dependency configuration rather than evidence that the current sentvia-mcp package is malicious.

Attack Path

  1. An attacker compromises the npm publisher account, package, registry delivery path, or a relevant dependency.
  2. The attacker publishes a malicious release under the existing sentvia-mcp package name.
  3. An operator follows the documented setup command.
  4. npx -y retrieves the package selected by npm's unpinned version resolution and executes it without an interactive confirmation.
  5. The executed package runs with the privileges and environment of the agent process.
  6. Because the instructions require placing SENTVIA_API_KEY in the MCP server environment, malicious package code could read and exfiltrate that credential.
  7. The attacker could then use the exposed credential to perform email operations allowed by the associated SentVia account.

Impact Assessment

Successful exploitation could provide arbitrary code execution with the privileges of the user running OpenClaw or npx. The accessible scope may include:

  • The SENTVIA_API_KEY supplied to the MCP server.
  • SentVia ...[truncated 383 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin sentvia-mcp to a reviewed, immutable version rather than relying on npm's default version resolution.
  2. Maintain a lockfile and verify the package and transitive-dependency integrity hashes.
  3. Document the official npm package URL, expected publisher identity, and approved version.
  4. Remove -y where practical so package installation is not silently approved.
  5. Install dependencies during a controlled deployment phase rather than dynamically at skill invocation time.
  6. Review new package versions before updating the approved pin.
  7. Run the MCP server in a sandbox or dedicated low-privilege account with restricted filesystem and network access.
  8. Expose only SENTVIA_API_KEY to the MCP process and avoid inheriting unrelated credentials or sensitive environment variables.
  9. Restrict the SentVia credential to the minimum permissions required and rotate it if package compromise is suspected.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (11)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The skill description is broadly phrased and does not state clear activation constraints or user-approval requirements before sending or managing email. In an agent ecosystem, this can cause over-broad invocation for tasks involving communication, increasing the chance of unintended outbound actions or interaction with untrusted external parties.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill enables creation of live inboxes and transmission of real email to external recipients, but it does not present a prominent user-facing warning that content and metadata will leave the local environment. This raises the risk of accidental disclosure of sensitive data, especially because the skill is designed for routine external communication and mailbox handling.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The skill declares a real external API endpoint and instructs use of a bearer token from environment variables, which is a clear external data transmission capability. In context this is intentional functionality, but it is still security-relevant because any misuse of the skill can send data off-system and mishandling of the token could expose mailbox control.

Content

Scanner excerpt · SKILL.md (reported line 18)May include surrounding context.

md
---

You can send and receive real email through SentVia. Your API key is in
`SENTVIA_API_KEY`; all requests go to `https://api.sentvia.ai/v1` with
`Authorization: Bearer $SENTVIA_API_KEY` and JSON bodies.

If the SentVia MCP server is configured (`sentvia` in `mcp.servers`), prefer its

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The skill instructs users to install and run an MCP server via npx ... sentvia-mcp without pinning a specific version or integrity hash. That creates a supply-chain risk: a future malicious or compromised package release would be fetched and executed automatically, potentially exposing the SentVia API key or enabling arbitrary code execution in the agent environment.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

This section operationalizes raw HTTP use for a third-party email service, enabling direct external transmission of agent-generated content and metadata. While expected for an email skill, it increases exfiltration and unintended-contact risk because the skill can act on real-world communications with only general safety guidance.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

md
tools (`create_inbox`, `send_message`, `reply_to_message`, `search_messages`, …)
over raw HTTP. To set it up:
`openclaw mcp add sentvia --command npx --arg -y --arg sentvia-mcp` and put
`SENTVIA_API_KEY` in that server's `env`. Otherwise use curl as below.

## Your address

External Transmission

Medium
Category
Data Exfiltration
Confidence
85% confidence
Finding

Listing inboxes against the external service transmits authentication credentials and retrieves mailbox metadata from a third party. Although this is core functionality, it broadens the agent's ability to inspect communication channels and may expose account structure or identifiers if invoked without proper user awareness.

Content

Scanner excerpt · SKILL.md (reported line 32)May include surrounding context.

Create one inbox for yourself once, then reuse it. Check first:

bash
curl -s https://api.sentvia.ai/v1/inboxes -H "Authorization: Bearer $SENTVIA_API_KEY"

If you have none, create one (pick a short local part that fits your name):

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

Creating a new live inbox on an external provider expands the agent's external communications surface and can establish persistent new channels without strong user warning or approval requirements. That creates risk of unsanctioned communications, data collection, or long-lived mailboxes that outlast the intended task.

Content

Scanner excerpt · SKILL.md (reported line 38)May include surrounding context.

If you have none, create one (pick a short local part that fits your name):

bash
curl -s -X POST https://api.sentvia.ai/v1/inboxes \
  -H "Authorization: Bearer $SENTVIA_API_KEY" -H "Content-Type: application/json" \
  -d '{"local_part": "claw", "display_name": "Claw"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

This endpoint sends outbound email to arbitrary recipients, which is the highest-risk operation in the skill because it can directly exfiltrate sensitive information, contact external parties, or be abused for spam/phishing if the agent is manipulated. The context makes this more dangerous because the skill is specifically designed to operate a real mailbox and includes minimal hard gating for recipient or content approval.

Content

Scanner excerpt · SKILL.md (reported line 51)May include surrounding context.

Send (client_id makes retries idempotent — use one per logical send):

bash
curl -s -X POST https://api.sentvia.ai/v1/messages \
  -H "Authorization: Bearer $SENTVIA_API_KEY" -H "Content-Type: application/json" \
  -d '{"inbox_id": "<id>", "to": ["person@example.com"], "subject": "…", "text": "…", "client_id": "<unique>"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
84% confidence
Finding

Polling or searching messages retrieves real email content from an external service into the agent context, which can import untrusted text and potentially sensitive correspondence. The skill does note that inbound email is untrusted, which reduces but does not eliminate the risk of prompt injection, privacy exposure, or over-collection of mailbox data.

Content

Scanner excerpt · SKILL.md (reported line 59)May include surrounding context.

Check for new mail (no query returns recent; add query to search, mode=semantic to search by meaning):

bash
curl -s "https://api.sentvia.ai/v1/messages?limit=10" -H "Authorization: Bearer $SENTVIA_API_KEY"

**Reply — always reply to a message id, never compose a fresh email into an

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

Replying to an existing message sends externally and can be triggered in response to untrusted inbound content, making it particularly susceptible to social engineering or prompt-injection-driven actions. Even though the skill advises replying only by message ID for threading correctness, it does not require explicit approval before sending a response to an external party.

Content

Scanner excerpt · SKILL.md (reported line 66)May include surrounding context.

existing conversation** (threading headers are set for you):

bash
curl -s -X POST https://api.sentvia.ai/v1/messages/<message_id>/reply \
  -H "Authorization: Bearer $SENTVIA_API_KEY" -H "Content-Type: application/json" \
  -d '{"text": "…"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
78% confidence
Finding

Managing a server-side block list is an external state-changing action against a third-party service. It is lower risk than sending mail, but unauthorized or mistaken use could disrupt legitimate communications or be abused to suppress important senders.

Content

Scanner excerpt · SKILL.md (reported line 79)May include surrounding context.

Block a noisy sender (enforced server-side, both directions):

bash
curl -s -X POST https://api.sentvia.ai/v1/lists \
  -H "Authorization: Bearer $SENTVIA_API_KEY" -H "Content-Type: application/json" \
  -d '{"kind": "block", "pattern": "spammy.example"}'