The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands git stash and git stash pop to manage branch states. These are standard development operations used to switch between 'before' and 'after' states of a codebase.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a browser (browser_navigate) which then processes the content of the page for screenshotting and description.
Ingestion points: The browser_navigate tool is instructed to visit http://localhost:3000/affected-page (SKILL.md).
Boundary markers: Absent; there are no specific instructions to the agent to ignore potentially malicious text within the browser's rendered content.
Capability inventory: The skill has access to shell commands (git) and workspace file writing (saving screenshots).
Sanitization: Absent; the visual content is passed directly to the agent's vision/context.
Contextual Risk: The risk is assessed as safe because the tool is explicitly designed for a developer to use on their own local development server, meaning the content being 'ingested' is the developer's own code.

screenshotting-changelog