retain
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to collect conversation-derived content, including summaries, concepts, and generated exercises, and transmit them to the Vidbyte service (vidbyte.pro) via a CLI command. This behavior is the primary stated purpose of the skill.\n- [COMMAND_EXECUTION]: The skill generates and executes a local shell command (
vidbyte retain) with multiple flags. It uses the CLI tool to transport structured data from the agent to the service backend.\n- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it processes untrusted conversation history and interpolates it directly into command-line arguments without instructions for sanitization.\n - Ingestion points: Current conversation history (SKILL.md).\n
- Boundary markers: Absent; the instructions do not specify the use of delimiters or 'ignore' warnings for the extracted conversation content.\n
- Capability inventory: Execution of local shell commands via the
vidbyteCLI tool.\n - Sanitization: Absent; strings extracted from the conversation are passed directly into CLI flags as variables like
$CONCEPT_1_NAME.\n- [EXTERNAL_DOWNLOADS]: The skill includes instructions to install thevidbyte-skillspackage from the npm registry if the required CLI binary is not present on the system.
Audit Metadata