The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches data from external URLs (e.g., hnrss.org, reddit.com, github.com) and processes the content (titles, links, descriptions) to be read by the agent. This represents an attack surface where an attacker could place malicious instructions in an RSS feed item to influence the agent's subsequent actions.
Ingestion points: The curl command is used throughout SKILL.md to fetch data from various remote RSS/Atom feeds.
Boundary markers: None. The instructions do not define delimiters or warnings to ignore instructions embedded in the fetched feed content.
Capability inventory: The skill uses curl for network requests and utilities like xmllint, grep, awk, and tr for text processing.
Sanitization: The skill performs structural parsing (XML/XPath) but does not sanitize the resulting natural language text for potential prompt injection patterns.
[REMOTE_CODE_EXECUTION]: An automated scanner flagged the pattern of downloading a file to /tmp/hn-feed.xml and then processing it. However, the tool used to process the file is xmllint, which is a static XML parser/querier, not a code execution engine. The content is treated as data, not as executable code.
[COMMAND_EXECUTION]: The skill utilizes several standard CLI tools to perform its tasks:
curl: Used to fetch remote feed data from well-known services.
xmllint: Used to extract specific XML nodes using XPath.
grep, awk, tr, head: Used for text filtering and formatting.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and content from well-known and reputable services including Hacker News (hnrss.org), TechCrunch, GitHub, and Reddit. These operations are neutral and consistent with the skill's stated purpose as an RSS reader.

rss-fetch