The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection due to its ingestion of untrusted external content.
Ingestion points: The agent reads the titles and context from the first 30 stories on the Hacker News front page (news.ycombinator.com).
Boundary markers: There are no boundary markers or specific instructions to ignore embedded directives within the processed data.
Capability inventory: The skill uses chrome-devtools to manipulate the browser, allowing it to enter text and click the 'Post' button on x.com automatically.
Sanitization: No sanitization or validation of the retrieved headlines is performed; the skill only filters for topic relevance.
[DATA_EXFILTRATION]: The skill performs network operations and interacts with active user sessions on external domains.
It accesses news.ycombinator.com to read data and x.com to publish content using a logged-in browser session.
Both domains are recognized as well-known technology services, and the interaction is consistent with the skill's stated primary purpose.

hn-to-x-poster