pitch-deck
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection where untrusted user content (pitch stories and slide lists) is interpolated into generated HTML and JavaScript code.
- Ingestion points: User-provided pitch stories and slide lists processed in Step 1 and Step 2 of
SKILL.md. - Boundary markers: None identified; user content is placed directly into HTML headers and body text.
- Capability inventory: The skill generates an HTML/JS file that executes in a browser environment, and provides a Python driver snippet that executes JavaScript via a separate recording skill (
demo-video). - Sanitization: The skill instructions in
references/deck-patterns.mddo not explicitly mandate escaping or sanitizing user-provided text before inserting it into the HTML skeleton. - [COMMAND_EXECUTION]: The skill instructs the user or agent to execute standard developer commands for local development and media processing.
- Evidence:
SKILL.mdandreferences/deck-recording.mddescribe usingpython3 -m http.serverfor local testing andffmpegfor stitching recorded video parts. - [EXTERNAL_DOWNLOADS]: The skill provides patterns for embedding external media into the deck.
- Evidence:
references/deck-patterns.mdincludes a template for embedding a YouTube video usingyoutube-nocookie.com. This is a well-known service and is documented here for completeness.
Audit Metadata