The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from GitHub Pull Requests to generate its storyboard and script. * Ingestion points: PR metadata and diff content are extracted into capture/extracted/visible-text.txt by ingest.mjs. * Boundary markers: No explicit security delimiters or 'ignore' instructions are added to the ingested content before it is processed by the agent. * Capability inventory: The agent has access to file system operations, network requests (via gh and fetch), and command execution (npx hyperframes). * Sanitization: Content is truncated and HTML comments are removed in ingest.mjs, but no semantic sanitization against malicious instructions is performed.
[COMMAND_EXECUTION]: The skill frequently executes shell commands via npx and the GitHub CLI (gh). * SKILL.md contains instructions to run npx hyperframes skills update, npx hyperframes init, and npx hyperframes render. * fetch-pr.mjs executes gh pr view, gh api, and gh pr diff.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from external services during operation. * fetch-people-avatars.mjs downloads contributor avatars from GitHub's infrastructure (github.com, *.githubusercontent.com). * captions.mjs and assemble-index.mjs include the GSAP animation library from cdn.jsdelivr.net in the generated HTML compositions. * audio.mjs interacts with the author's audio generation APIs.
[DATA_EXFILTRATION]: The skill reads data from GitHub repositories (PR details, diffs) and contributor profiles. While this is the intended purpose, it involves accessing potentially sensitive codebase information.

pr-to-video