krillinai-tts

Pass

Audited by Gen Agent Trust Hub on Oct 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the KrillinAI CLI tool through an environment variable $KRILLINAI_CLI within the directory $KRILLINAI_CWD. This is a standard integration for this vendor's ecosystem.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external SRT subtitle files, which provides an attack surface for indirect prompt injection if the subtitles contain malicious instructions meant for an agent processing the tool's output.
  • Ingestion points: The file target_language_srt.srt is ingested via the --input-srt flag in SKILL.md.
  • Boundary markers: None identified; standard SRT format is assumed.
  • Capability inventory: Executes a shell command using $KRILLINAI_CLI and performs file system operations via ffprobe.
  • Sanitization: No specific sanitization or validation of the SRT content is documented within the skill.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes a --voice-clone-source flag that accepts a URL. This allows the tool to perform network operations to fetch external audio content. While this is a functional requirement for voice cloning, it represents a network-access surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 27, 2026, 10:42 PM