The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted user-generated content from the Lobste.rs website, including story titles, descriptions, and comments.
Ingestion points: Data is retrieved from the Lobste.rs API via httpx.get requests in the fetch_stories, fetch_story, and search functions within scripts/lobsters.py.
Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions to isolate the fetched external data from the agent's primary instruction set, making the agent potentially susceptible to instructions embedded in story content.
Capability inventory: The skill uses httpx for network communications and typer for command-line interaction. It runs as a local Python script via uv run.
Sanitization: While scripts/lobsters.py includes an html_to_text helper to strip HTML tags and unescape entities, it does not perform semantic validation or filtering to detect or block prompt injection patterns within the retrieved text strings.

ez-lobsters