laniameda-instagram-carousel-extract
Warn
Audited by Gen Agent Trust Hub on Apr 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Potential shell command injection via the
<POST_URL>variable in the Supadata retrieval command. The URL is interpolated into acurlbash command; if the URL contains subshell execution syntax like$()or backticks, it could allow arbitrary command execution. - [PROMPT_INJECTION]: Vulnerability to indirect prompt injection from untrusted external data.
- Ingestion points: Instagram post captions and text visually transcribed from carousel slides by the browser agent.
- Boundary markers: Absent. No delimiters or instruction-isolation markers are used for the extracted text.
- Capability inventory: Execution of shell commands via
curland file-writing operations to the Knowledge Base. - Sanitization: Absent. The skill does not validate or sanitize extracted text before use in classification or storage.
- [EXTERNAL_DOWNLOADS]: Connects to the Supadata API to retrieve post metadata. This is standard functionality for the skill's data ingestion purpose.
Audit Metadata