The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted news signals from external sources (RSS feeds, X, and news searches) to generate PR reports and campaign prompts.
Ingestion points: Data retrieved from news_search and various RSS feeds listed in references/rss-feeds.json (e.g., Techmeme, Google News).
Boundary markers: The skill does not define specific delimiters for untrusted content, though it passes data to filtering and origin-checking sub-skills.
Capability inventory: The skill executes shell commands via the newsjack CLI, writes JSON and Markdown artifacts to the local filesystem, and can send messages to Slack webhooks.
Sanitization: No explicit content sanitization or escaping is mentioned before news summaries are included in reports or sent to Slack.
[COMMAND_EXECUTION]: The skill relies on a bundled CLI tool, newsjack, to perform core functions including story clustering, filtering, and triage.
Evidence: Multiple shell command examples in SKILL.md, such as newsjack detector run, newsjack filter-apply, and newsjack origin-apply.
[EXTERNAL_DOWNLOADS]: The skill retrieves news data and RSS feeds from various external domains.
Evidence: references/rss-feeds.json contains numerous URLs for news sources like Google News, Techmeme, and official government press releases (FTC, SEC, GOV.UK). These are well-known and reputable services.

newsjack-detector