scrollcraft

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill interacts with external APIs at api.kie.ai and kieai.redpandaai.co via the scripts/kie.mjs utility. These requests are used to upload local assets and download generated AI imagery and video. These domains are not included in the standard whitelist.- [COMMAND_EXECUTION]: Several local scripts (scripts/encode.sh, scripts/doctor.mjs, scripts/shoot.mjs) execute system commands to process video files using ffmpeg and perform browser-based visual verification using playwright. These are development-time tools intended for the build process.- [PROMPT_INJECTION]: The skill implements an interview-driven workflow where user-provided input is used to generate brief documents, art direction prompts for the AI models, and semantic HTML landing pages. This creates a surface for indirect prompt injection. * Ingestion points: User responses via AskUserQuestion and the BRIEF.md file (read in Step 2). * Boundary markers: Not present. * Capability inventory: File writing and editing (Write, Edit), shell script execution (Bash), and external API calls (scripts/kie.mjs). * Sanitization: No explicit sanitization or escaping of user-provided strings was detected before they are interpolated into prompts or HTML.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 08:59 PM
Security Audit — agent-trust-hub — scrollcraft