The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands to perform its duties. Specifically, it uses git to analyze local changes and the gh (GitHub) CLI to fetch pull request data, comments, and to post review summaries back to GitHub. It also executes a local Node.js script using the node command.
[EXTERNAL_DOWNLOADS]: Through the capture-screenshots.js script, the skill uses Playwright (Chromium) to navigate to external URLs, primarily on aem.page and aem.live domains. This is done to capture screenshots and perform visual validation of the changes. The skill relies on the well-known playwright package from the official NPM registry.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It ingests untrusted data from multiple sources including git diffs, GitHub PR comments, and the text content of the web pages it visits via Playwright.
Ingestion points: gh pr view, gh pr diff, gh api (comments), and the browser context in capture-screenshots.js (file: SKILL.md, scripts/capture-screenshots.js).
Boundary markers: The instructions do not define strict boundary markers or 'ignore' instructions when processing this external data.
Capability inventory: The skill can execute shell commands (git, gh, node), write files to the local disk, and access the network via a browser (file: SKILL.md, scripts/capture-screenshots.js).
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external sources before it is analyzed by the agent.

code-review