The Agent Skills Directory

[External Downloads]: The skill uses npm to install specific versions of @playwright/test for bisection and reproduction purposes. This involves fetching packages from the public npm registry. Since these packages are official components of the project being triaged, this is standard behavior for the skill's intended use case.
[Command Execution]: The skill executes shell commands such as git fetch, npm init, and npx playwright install. These are used to set up reproduction environments and investigate code history. The skill includes a clear instruction to avoid running these commands on pull requests that modify sensitive configuration or infrastructure files, such as package.json or .github/ workflows, which serves as a protective boundary.
[Indirect Prompt Injection Surface]: Because the skill processes external data from GitHub issues (such as reproduction steps or bug descriptions), there is a theoretical surface for indirect prompt injection. However, the skill is scoped to triage tasks and provides explicit instructions on which files and PR types are safe to interact with, reducing the likelihood of unexpected behavior.
[Dynamic Environment Creation]: The skill dynamically creates scratch directories to test different versions of the software. This is a functional requirement for its bisection capabilities and is managed through standard file system operations.

playwright-triage