song-ad
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to perform local file system operations and media processing using shell commands. \n
- Evidence: In
timing-and-slicing.md, the agent is provided with command templates forffmpeg(audio conversion, trimming, and slicing),ffprobe(duration measurement),stat(file size checking), andcurl(downloading source files and uploading processed slices). \n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external data from the brand's profile and catalog. \n - Ingestion points: The skill pulls brand briefs, dossiers, and product catalogs using
advibly_get_brand,advibly_get_brand_dossier, andadvibly_get_products(SKILL.md). \n - Boundary markers: Absent. The instructions do not specify the use of delimiters or "ignore" warnings for the ingested brand data. \n
- Capability inventory: The agent has the ability to execute shell commands (
ffmpeg,curl), perform file system operations, and invoke various generative AI tools for music, image, and video creation. \n - Sanitization: Absent. There are no instructions for sanitizing or validating the content of the brand data before it is interpolated into prompts. \n- [EXTERNAL_DOWNLOADS]: The skill fetches generated audio assets from the vendor's platform. \n
- Evidence: The skill uses
curlto download song takes from URLs provided by theadvibly_get_generationtool for local processing intiming-and-slicing.md.
Audit Metadata