deck-design-ppt
Warn
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The pattern loading logic in
masters/index.jsuses dynamic string interpolation in arequirecall (require("./patterns/${name}")). This allows for potential path traversal if the pattern name is influenced by an subverted agent or external input, which could lead to the execution of arbitrary JavaScript files within the environment. - [PROMPT_INJECTION]: The skill's operational algorithm described in
SKILL.mdinvolves analyzing and processing user-supplied briefs and structured data to build slide content. This creates an attack surface for indirect prompt injection, where malicious instructions embedded in the untrusted input data could influence the agent's behavior during the deck generation process. - [DATA_EXFILTRATION]: The utility script
masters/html2pptx.jsuses a headless browser to render HTML content from local file paths. If this utility were to be used with untrusted file paths, it could potentially be exploited to access and capture screenshots of sensitive local system files, which could then be embedded in the generated presentation deliverable.
Audit Metadata