create-product-images-higgsfield-product-photoshoot

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied briefs and source assets which are interpolated into prompts for external generative AI models.\n
  • Ingestion points: User brief and source asset paths defined in SKILL.md.\n
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are documented for the prompt construction.\n
  • Capability inventory: The skill uses file writing, network calls to AI APIs, and shell-based media processing via internal scripts.\n
  • Sanitization: Includes logic for scrubbing NSFW vocabulary from prompts as described in references/nsfw_routing.md.\n- [DYNAMIC_EXECUTION]: The skill uses scripts/render_hyperframe.py to execute HTML and GSAP (JavaScript) code via Playwright for rendering pixel-perfect graphics.\n
  • The logic involves frame-by-frame rendering driven by document.timeline.currentTime.\n- [COMMAND_EXECUTION]: The skill documentation references several internal Python scripts (scripts/composite.py, scripts/make_subtitles.py, scripts/qc_review.py) that perform shell-level operations using ffmpeg and playwright for video assembly and quality control.\n- [EXTERNAL_DOWNLOADS]: The skill documentation suggests environment setup steps including pip install playwright pillow and system package installations (brew install ffmpeg-full or apt install ffmpeg). It also makes network requests to various AI service providers (fal.ai, ElevenLabs, Veed, Higgsfield). These are documented as necessary dependencies for the media generation task.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 04:39 PM
Security Audit — agent-trust-hub — create-product-images-higgsfield-product-photoshoot