bibi-feed
Fail
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill suggests installing the
bibiCLI by piping a remote script fromhttps://bibigpt.co/install.shto bash, which is an execution pattern for remote code from the vendor's site. - [EXTERNAL_DOWNLOADS]: The skill uses external resources from vendor-controlled domains
bibigpt.coandapi.bibigpt.cofor tool installation and API communication. - [COMMAND_EXECUTION]: The skill relies on executing the
bibicommand-line utility for its core workflows. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted metadata from external video platforms, presenting a potential surface for injection.
- Ingestion points: Video titles and descriptions retrieved via
bibi channels videosandbibi feedin the channel management and feed workflows. - Boundary markers: Absent; there are no instructions to use delimiters or ignore instructions within the ingested metadata.
- Capability inventory: Shell command execution via the
bibiCLI. - Sanitization: Absent; external data is processed and presented without specific filtering or escaping.
Recommendations
- HIGH: Downloads and executes remote code from: https://bibigpt.co/install.sh - DO NOT USE without thorough review
Audit Metadata