ppt-avatar-video
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted presentation files (PPT, PPTX, PDF) to extract slide text and speaker notes, which are then used to generate narration scripts.
- Ingestion points: The skill downloads files using
okou web download-fileand extracts content viaokou presentation screenshotand manual parsing of deck data. - Boundary markers: Narration text is handled as file content and piped through standard input to generation tools, reducing but not eliminating the risk of command-like strings being interpreted by the narration provider.
- Capability inventory: The skill possesses significant local capabilities including file system write access, execution of local scripts, and network rendering via cloud APIs.
- Sanitization: The
scripts/build-project.mjsscript includes anescapeHtmlfunction to prevent malicious text in slides from breaking the generated HTML composition project. - [COMMAND_EXECUTION]: The skill relies heavily on executing shell commands and local Node.js scripts to perform media processing and project assembly.
- Evidence:
SKILL.mdcontains numerous templates forokou,node, andnpxcommands. The helper scripts usenode:child_processto invokeffmpegandffprobefor media analysis. - [EXTERNAL_DOWNLOADS]: The skill downloads and executes an external package from the NPM registry at runtime.
- Evidence:
SKILL.mdandscripts/build-project.mjsboth invokenpx --yes hyperframes@0.8.26to perform project validation and rendering tasks. - [DYNAMIC_EXECUTION]: The skill generates executable HTML and configuration files at runtime to define the video composition.
- Evidence: The
scripts/build-project.mjsscript programmatically constructs anindex.htmlfile using slide paths and metadata, which is then 'executed' by the Hyperframes renderer to produce the final video.
Audit Metadata