Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward Google-backed audio transcription tool, with remote upload and Google credential use aligned with its stated purpose.
Install this only if you are comfortable sending the selected audio files to Google Gemini or Vertex AI and using your Google API key or gcloud credentials for the request. Avoid transcribing regulated, confidential, or highly sensitive recordings unless that matches your organization's data-handling rules.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
def get_gcloud_access_token() -> str | None:
"""Get access token from gcloud CLI (Application Default Credentials)."""
try:
result = subprocess.run(
["gcloud", "auth", "print-access-token"],
The skill advertises code-backed behavior that accesses environment variables, invokes shell-oriented setup commands, and performs outbound network requests, but it declares no explicit tool scope or permissions boundary. In an agent ecosystem, this weakens user visibility and policy enforcement, increasing the chance that sensitive environment data or external network access is used without informed approval.
The documentation explains transcription but does not clearly warn that the provided audio is uploaded to Google Gemini API or Vertex AI for remote processing. This can cause users or downstream agents to send sensitive voice recordings, background conversations, or regulated data off-device without informed consent, creating privacy, compliance, and data-handling risk.
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
def get_gcloud_access_token() -> str | None:
"""Get access token from gcloud CLI (Application Default Credentials)."""
try:
result = subprocess.run(
["gcloud", "auth", "print-access-token"],
capture_output=True,
text=True,
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
# Fall back to gcloud config
try:
result = subprocess.run(
["gcloud", "config", "get-value", "project"],
capture_output=True,
text=True,
This function reads the full local audio file, base64-encodes it, and sends it to Google's Gemini API for remote processing, but the runtime behavior does not provide an explicit user-facing warning or consent checkpoint. In a transcription skill, remote transmission is functionally expected, but sending potentially sensitive voice data off-host without clear disclosure can create privacy, compliance, and data-handling risk.
The Vertex AI path transmits user audio to a remote Google endpoint and includes a bearer token for authorization, again without an explicit user-facing disclosure at the point of use. Although HTTPS is used and this is aligned with the stated feature set, the absence of a clear warning or consent mechanism increases privacy and enterprise policy risk when handling sensitive recordings.
No suspicious patterns detected.