The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, including GitHub PR titles, bodies, and diffs, as well as marketing briefs and blog posts to derive video narratives.
Ingestion points: Steps 1.2 and 1.2a in SKILL.md retrieve data via gh pr view and file reads.
Boundary markers: The instructions do not specify the use of delimiters or "ignore embedded instructions" warnings when passing this data to the agent's internal reasoning steps.
Capability inventory: The skill can execute shell commands (npx, gh, git, pnpm), write to the filesystem, and open local URLs in a browser.
Sanitization: There is no evidence of filtering or escaping logic for the ingested natural language text before it is used to derive motifs, hooks, or scene plans.
[COMMAND_EXECUTION]: The skill relies on several shell commands to perform its core functions.
It uses gh pr view and gh pr diff for GitHub integration.
It uses git diff and git log for local repository analysis.
It executes npx hyperframes for project initialization, previewing, and rendering.
It uses system-level commands (open, xdg-open, start) to launch the preview interface in a browser.
[EXTERNAL_DOWNLOADS]: The skill fetches assets and dependencies from well-known and reputable services.
It installs the shiki package using the project's detected package manager.
It references the gsap library from the jsdelivr.net CDN within HTML templates.
It loads typography from fonts.googleapis.com.
[DYNAMIC_EXECUTION]: The skill dynamically generates HTML compositions and GSAP-based animation scripts from templates and project data.
It writes these generated files to the local filesystem and executes them through a local preview server and browser rendering engine.
This behavior is the primary purpose of the skill and is constrained to the project output directory.

hyperframes-video