The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to generate JavaScript scenario files in a temporary directory (/tmp) and subsequently executes them using the Node.js import() function in scripts/prepare-demo.mjs and scripts/record-demo.mjs. This mechanism allows the agent to define and run custom browser automation scripts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external, potentially untrusted data such as pull request diffs and test configurations to plan and record demonstrations, which represents a vulnerability surface for instructions embedded in code or configuration.
Ingestion points: Pull request diffs and test configurations are read as part of the feature inspection step in SKILL.md.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the diff content.
Capability inventory: The skill possesses the ability to execute shell commands (ffmpeg, ffprobe, gh), write files to the system, and perform network operations via Playwright browser automation and the GitHub CLI.
Sanitization: No sanitization or filtering of the ingested pull request content is performed before the agent processes it.
[COMMAND_EXECUTION]: The skill frequently executes shell commands to process media and interact with GitHub.
Evidence: scripts/convert-demo.sh invokes ffmpeg and ffprobe. scripts/prepare-demo.mjs uses spawnSync to call the conversion script as well as ffmpeg and ffprobe directly. The publishing workflow in references/publishing.md utilizes the gh (GitHub CLI) to view, edit, and comment on pull requests.
[EXTERNAL_DOWNLOADS]: The skill relies on the @playwright/test package, which is designed to download and execute browser binaries (e.g., Chromium) to facilitate the recording process.

record-pr-demo