The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied Remotion source files for translation to HyperFrames. This creates a surface for indirect prompt injection if the source code contains malicious instructions in comments or string literals intended to influence the AI agent's behavior during the migration process. However, this risk is inherent to the primary function of code translation and is managed by the agent's narrow task scope. Sanitization is limited to compatibility checks performed by the provided linter.\n- [COMMAND_EXECUTION]: The skill utilizes several local shell scripts, such as scripts/render_diff.sh and assets/test-corpus/run.sh, to orchestrate the video rendering and comparison pipeline. These scripts execute standard command-line utilities including ffmpeg, ffprobe, npm, and npx. The usage is transparent and directly supports the tool's intended functionality.\n- [EXTERNAL_DOWNLOADS]: The skill performs package installation via npm install for its test fixtures and references external JavaScript libraries including GSAP and Lottie-web from well-known content delivery networks (cdnjs.cloudflare.com, unpkg.com). The dependencies used, such as React and Remotion, are standard, legitimate software packages.\n- [DYNAMIC_EXECUTION]: The evaluation harness uses a common pattern of embedding Python logic within bash scripts via heredocs (`python3
<<'PY' ...`). This dynamic execution is local to the skill's scripts and is used to perform technical calculations like SSIM metrics and frame sampling during the render validation process.

remotion-to-hyperframes