Security audit
Security checks for vulnerabilities and agentic risk
This is a cloud transcription skill, but its URL handling is too broad and could upload unintended local or internal data to the transcription provider.
Review before installing. Use this only if you are comfortable sending selected audio, filenames, and downloaded URL contents to SkillBoss. Do not give it sensitive recordings, private/internal URLs, localhost URLs, cloud metadata URLs, or very large files unless the publisher adds URL validation, media and size limits, JSON-safe request construction, and clearer data-handling disclosure.
scripts/asr.sh:43Unrestricted URL Retrieval Can Disclose Local or Internal Data
scripts/asr.sh:59Unescaped User-Controlled Values Permit JSON Request Injection
scripts/asr.sh:43Unbounded Download and Base64 Processing Enable Resource Exhaustion
The skill explicitly supports sending either a URL or a local file to a third-party ASR service, but the documentation does not warn users that local media contents and referenced remote resources will be transmitted off-host. This can cause unintentional disclosure of sensitive audio, embedded personal data, or internal-only URLs in automated agent workflows where users may assume processing is local.
The manifest declares an external API credential (SKILLBOSS_API_KEY) and describes cloud-based audio transcription, but it does not warn users that their audio may be transmitted to a third-party service for processing. This creates a meaningful privacy and consent risk because users may provide sensitive recordings without clear disclosure that data leaves the local environment and is handled by an external provider.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
#!/usr/bin/env bash
# ASR CLI Wrapper powered by SkillBoss API Hub
# STT endpoint: https://api.skillboss.co/v1/pilot
API_BASE="https://api.skillboss.co/v1"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
#!/usr/bin/env bash
# ASR CLI Wrapper powered by SkillBoss API Hub
# STT endpoint: https://api.skillboss.co/v1/pilot
API_BASE="https://api.skillboss.co/v1"
The script will fetch any user-supplied URL and upload the retrieved content to a third-party API, creating a server-side data transfer capability that can be abused to access internal or sensitive resources if the environment running the skill has privileged network access. Even though this appears intended to support remote audio transcription, unrestricted URL fetching expands the trust boundary and can expose local-network services or unexpected data to exfiltration.
This line performs the actual transmission of collected audio data and filename metadata to an external service using an API key, which is a real data egress point. In the context of an ASR skill this may be functional, but it is still security-relevant because sensitive audio can leave the local environment and be processed by a third party.
}
EOF
curl -s -X POST "$API_BASE/pilot" \
-H "Authorization: Bearer $SKILLBOSS_API_KEY" \
-H "Content-Type: application/json" \
-d @"$TMPREQ"
The skill transmits user-provided audio and metadata to an external API but does not provide an explicit privacy or upload warning at the point of use. In an agent skill context, this matters because users may assume local processing, while the script actually exfiltrates potentially sensitive speech content to a third party.
The documentation states that STT is synchronous and that results are returned immediately, yet the code accepts flags commonly associated with asynchronous or streaming behavior without implementing them. This is an active mismatch between documented behavior and code interface because users are led to believe these options are meaningful when they are discarded.
No suspicious patterns detected.