scenes-gathered-zine-live-flow-v1-0
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The orchestration script
scripts/build_live_zip_and_import.pyusessubprocess.runto execute internal logic and system tools. - It searches for the importer application using
mdfindto ensure the correct tool is used. - It executes
scripts/render_adaptive_flat_tear.pyvia the Python interpreter to generate animation frames. - It invokes the Swift interpreter to run
scripts/pair_live_photo.swiftfor metadata processing. - It uses the macOS
opencommand to pass the completed ZIP file to the bundled实况照片导入器.app. - [COMMAND_EXECUTION]: The rendering script
scripts/render_adaptive_flat_tear.pycallsffmpegto encode individual images into a QuickTime MOV file. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of images and user-supplied coordinate strings.
- Ingestion points: The
--inputimage path and the--path-pointsstring argument inscripts/build_live_zip_and_import.py. - Boundary markers: The coordinate strings lack explicit boundary delimiters, but the rendering engine applies strict parsing rules.
- Capability inventory: The skill has the ability to write to the file system, execute local scripts and system binaries, and invoke external macOS applications.
- Sanitization: In
scripts/render_adaptive_flat_tear.py, thepath-pointsstring is split and each component is explicitly converted to a float, which effectively prevents command injection via that specific parameter.
Audit Metadata