The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external URLs provided as arguments, which creates an attack surface for indirect prompt injection.
Ingestion points: Content is ingested from the target web page in SKILL.md via the command playwright-cli open "$URL" across four different testing agents.
Boundary markers: The instructions do not define explicit boundary markers, delimiters, or "ignore embedded instructions" directives for the content retrieved from the target web page.
Capability inventory: The skill utilizes playwright-cli to execute JavaScript (eval, run-code) within the browser context of the target page and writes results (screenshots, snapshots, and traces) to the local file system using the mkdir and playwright-cli tools.
Sanitization: There is no evidence of sanitization, filtering, or validation of the content retrieved from the web page before it is processed by the agent or included in the final markdown report.

test-a11y