krillinai-cli
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill directs the agent to compile the KrillinAI CLI binaries at runtime from the local repository using
pnpm krillinai:build. This involves executing build scripts to generate native binaries (krillinai-cliandkrillinai-server) based on the host platform and architecture. - [COMMAND_EXECUTION]: The agent is instructed to execute the compiled
krillinai-clibinary with various flags and subcommands (e.g.,subtitle,tts,render-vertical). This includes setting environment variables likeOPENCREATOR_KRILLINAI_CLIand managing local configuration files (config/config.toml). - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process external media data, such as YouTube videos and subtitle files, which introduces a potential surface for indirect prompt injection if the processed content contains adversarial instructions.
- Ingestion points: The skill ingests untrusted data from external sources, specifically YouTube URLs (e.g.,
https://www.youtube.com/watch?v=VIDEO_ID) and local media files provided as inputs to the CLI. - Boundary markers: No specific boundary markers or instructions to disregard embedded commands in processed data are defined in the instructions.
- Capability inventory: The skill possesses extensive local capabilities, including shell command execution (
pnpm,node,krillinai-cli) and the ability to write files to the filesystem (--workdir). - Sanitization: There is no evidence of explicit sanitization or filtering of the content retrieved from external media before it is processed by the CLI or presented to the agent context.
Audit Metadata