stickman-video
Warn
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
assets/example/tighten-pauses.mjscontains a command injection vulnerability. It uses a helper function that executes shell commands viash -cand interpolates theinpandoutpvariables (taken from command-line arguments) directly into the command string forffmpegandffprobe. This allows for arbitrary command execution if filenames contain shell metacharacters like backticks or subshell markers. - [EXTERNAL_DOWNLOADS]: The skill uses
npxandbunxto download and execute thehyperframesutility from external registries. It also employscurlto interact with thevarg.aiAPI for voiceover generation, credit management, and balance checks. - [DYNAMIC_EXECUTION]: The
assets/example/sfx.mjsscript performs procedural audio synthesis usingffmpegfilters likeaevalsrcandanoisesrc, which evaluate mathematical expressions at runtime to generate audio waveforms dynamically. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from
assets/audio/transcript.jsonand interpolates it directly intoindex.htmlvia thecaptions.mjsscript. It lacks sanitization or boundary markers, creating a vulnerability surface where malicious content in a voiceover transcript could inject arbitrary HTML or influence the agent's behavior when it subsequently processes the file.
Audit Metadata