wjs-voicedrop-reading-aloud
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands and Python scripts for audio synthesis and validation. It specifically uses
python3 ~/code/volcano-tts/tts.pyto generate audio andafinfoto verify the resulting MP3 files. It also references a transcription script located at~/.claude/skills/wjs-transcribing-audio/scripts/volc_asr_stream.pyfor quality control. - [EXTERNAL_DOWNLOADS]: The skill is designed to fetch content from external URLs using web retrieval tools and browser skills to obtain text for narrating. It also includes references to official Volcengine documentation for voice configuration.
- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it ingests and processes untrusted data from URLs, files, and third-party articles.
- Ingestion points: Content is retrieved from user-provided text, files, external URLs via WebFetch/browse, and VoiceDrop articles.
- Boundary markers: The instructions focus on rewriting and reformatting content for TTS quality but do not specify security delimiters to isolate untrusted input from the agent's instructions.
- Capability inventory: The skill has the ability to execute shell commands, read environment variables from a local file, and write generated audio files to the local file system.
- Sanitization: The skill removes visual artifacts (markdown, links, etc.) to improve narration quality, but does not implement specific sanitization to prevent adversarial instructions within the input from influencing the agent's behavior.
Audit Metadata