The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of fetching and summarizing untrusted external data.
Ingestion points: The scripts/aggregate.py script reads content from a large list of external URLs defined in references/feeds.opml using the feedparser library.
Boundary markers: The script outputs summarized feed content using standard Markdown (e.g., ### Title and - **Summary**: {summary}). It lacks robust delimiters or explicit instructions for the agent to treat the summarized content as untrusted data or to ignore any embedded instructions.
Capability inventory: The skill has the capability to perform network requests (GET) to fetch RSS content and displays the results to the agent. It does not perform file-writes or execute shell commands based on the fetched content.
Sanitization: The clean_summary function in scripts/aggregate.py uses regex (re.sub(r'<[^>]+>', '', summary)) to strip HTML tags and unescapes HTML entities. While this cleans the presentation, it does not sanitize the underlying natural language text to prevent prompt injection or instruction overrides.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data from numerous external domains listed in references/feeds.opml. While this is the intended purpose of the skill, the wide range of third-party sources increases the surface area for processing potentially malicious or deceptive content.

rss-aggregator