T09 · Insecure Skill Coding Practices
- Location
index.js:20- Finding
API Credentials Can Be Redirected to an Arbitrary or Plaintext Endpoint
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill appears to be a legitimate Sayba marketplace MCP server, but it should be reviewed because it handles account credentials and public marketplace actions with weak endpoint and installation safeguards.
Install only if you trust the Sayba package and the configured marketplace endpoint. Prefer pinning the npm package to a reviewed version, use the default HTTPS Sayba host unless you intentionally operate a custom instance, avoid reusing a production API key with custom hosts, and confirm before letting an agent publish skills, submit reviews, invoke paid skills, or send sensitive inputs.
index.js:20API Credentials Can Be Redirected to an Arbitrary or Plaintext Endpoint
SKILL.md:39Unpinned npm Package Is Automatically Downloaded and Executed
package-lock.json:27Dependency Lockfile Resolves Packages Through a Third-Party Registry Mirror
fast-uri 3.1.2 is flagged for multiple host-parsing and normalization flaws, including SSRF-relevant confusion cases. In an MCP marketplace/server context that may process user-supplied URLs, URI parsing ambiguities can materially weaken allowlists, SSRF defenses, and origin validation.
hono 4.12.21 has numerous advisories, including CORS misconfiguration, proxy/header handling issues, and DoS conditions. Because this skill is an MCP server/marketplace component that likely handles inbound network traffic, weaknesses in the web framework are especially relevant and increase operational exposure.
ip-address 10.2.0 is affected by multiple parsing and classification issues that can undermine SSRF protections and network boundary checks. In a marketplace/server skill, any feature that validates client IPs, proxy addresses, callback targets, or internal-network restrictions could be bypassed if it relies on this library's incorrect interpretations.
The README advertises tools that publish skills, invoke remote skills, and retrieve call history using an API key, but it does not clearly warn that these operations transmit data to an external service and may expose prompts, inputs, metadata, or account-linked activity. In an MCP context, users may install and grant credentials based on the README alone, so missing disclosure increases the risk of unintended data sharing and unexpected authenticated actions.
The skill exposes capabilities that can access environment variables and the network, but it does not declare any explicit tool scope such as permissions or allowed-tools. This weakens least-privilege controls and can cause users or hosting platforms to grant broader access than necessary, increasing the risk of API key exposure or unintended outbound requests.
npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.
npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.
The publish_skill tool performs an external state-changing operation that may publish content to a marketplace immediately, yet the tool provides no built-in confirmation gate, dry-run mode, or explicit warning that the action can make content publicly available. In an MCP setting, an agent could invoke this tool based on ambiguous user intent or prompt manipulation, causing unintended publication of prompts, schemas, or proprietary material.
The rate_skill tool sends user-supplied review text to an external service without an explicit disclosure at invocation time that the text will leave the local environment and be stored remotely. In MCP/agent workflows, that can lead to accidental exfiltration of sensitive or private text if the agent composes or forwards content without the user's informed consent.
The lockfile pins @hono/node-server 1.19.14, and the cited advisory describes a Windows-specific path traversal in Hono static file serving. Although a lockfile alone does not prove the vulnerable code path is used, inclusion of the affected package is a real supply-chain risk because this skill depends on an SDK that may expose HTTP serving behavior.
body-parser 2.2.2 is flagged for a denial-of-service issue involving invalid limit handling. This is a real dependency risk, though exploitability depends on whether the skill exposes Express body parsing on attacker-controlled requests through the SDK stack.
qs 6.15.2 is flagged for denial-of-service and parsing-bypass issues. This is a genuine dependency concern, though practical impact depends on whether untrusted query strings or form bodies are parsed through Express/router paths exposed by the skill.
The dependency is version-ranged with a caret, which allows automatic installation of newer minor/patch releases that may introduce malicious code, regressions, or supply-chain compromises without explicit review. In an MCP server, dependency compromise can affect all server behavior and potentially expose data handled by the skill, though this package file alone does not show more severe amplifying behavior.
},
"homepage": "https://ai.sayba.com",
"dependencies": {
"@modelcontextprotocol/sdk": "^1.12.0"
}
}