The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from various external sources to generate PR briefs, newsjacking angles, and podcast prep notes, which creates an attack surface for indirect prompt injection.
Ingestion points: references/newsjacking.md (fetches content from Google News RSS, Hacker News Algolia API, and Reddit JSON) and references/podcast-guest-prep.md (fetches RSS feeds, Apple Podcasts descriptions, and general web search results).
Boundary markers: Absent. The instructions do not define delimiters or provide specific warnings to the agent to treat the external content as untrusted data.
Capability inventory: The skill is limited to text generation and research; no dangerous system-level capabilities such as eval, exec, or arbitrary file-writing were identified in the scripts.
Sanitization: Absent. There are no instructions for sanitizing, escaping, or filtering the external content before it is interpolated into the agent's context.
[EXTERNAL_DOWNLOADS]: The skill fetches data from several external services to provide real-time news updates and research.
Fetches news headlines and descriptions from Google News RSS feeds.
Retrieves trending tech stories and metadata from the Hacker News API hosted by Algolia.
Downloads JSON data from specific subreddits to monitor community trends.
Accesses podcast RSS feeds and Apple Podcasts show pages to extract episode metadata and descriptions.
[COMMAND_EXECUTION]: The skill provides shell commands intended for the agent or user to gather research data and prepare the environment.
Includes curl commands to interact with news APIs and RSS endpoints.
Uses jq and xmllint for parsing the downloaded structured data.
Recommends the use of brew install jq to ensure necessary tools are available for data processing.

public-relations