The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically generates a JavaScript file (/tmp/highlight.js) and a JSON configuration file (/tmp/shots.config.json) using cat and printf commands. These files are subsequently executed or loaded by the playwright-cli tool to manipulate the browser environment and capture screenshots.
[INDIRECT_PROMPT_INJECTION]: The skill operates by opening external URLs and processing their content to generate visual evidence, which the agent is then instructed to review before posting. This creates a potential surface where malicious content on a web page could attempt to influence the agent's judgment during the manual review phase.
Ingestion points: Web page content rendered via playwright-cli and captured in screenshots.
Boundary markers: None present for the visual content; the skill relies on a manual checklist for the agent to follow.
Capability inventory: Browser interaction and evaluation (playwright-cli), file system writes (cat, printf), and remote attachment/upload (gh-attach).
Sanitization: No sanitization is performed on the web content before it is visually presented to the agent for verification.

screenshots