The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from various sources, which presents a surface for indirect prompt injection attacks. Ingestion points: In references/newsjacking.md, the skill fetches content from Google News RSS, Hacker News (via Algolia API), and Reddit. In references/podcast-guest-prep.md, it processes RSS feeds and podcast website descriptions. Boundary markers: The skill instructions do not specify any delimiters or safety warnings to ignore instructions embedded within the fetched external content. Capability inventory: The skill uses shell commands (curl, jq, xmllint) in references/newsjacking.md and browser-driven research in references/journalist-pitching.md and references/podcast-guest-prep.md. Sanitization: There is no evidence of sanitization, validation, or filtering applied to the external data before it is processed by the agent.
[COMMAND_EXECUTION]: The skill includes shell command templates for retrieving data from the web. Evidence: references/newsjacking.md contains bash snippets using curl to fetch data from well-known domains (news.google.com, hn.algolia.com, reddit.com) and pipes the output to jq or xmllint for parsing. While targeting established services, these commands represent the capability to process arbitrary external data via the shell.

public-relations