The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network requests to www.saos.org.pl via a Python script to fetch legal judgments. While this is the intended functionality, it involves communication with an external domain not on the predefined whitelist.
[PROMPT_INJECTION]: The skill processes text from Polish court judgments retrieved from an external API, which creates a surface for indirect prompt injection where instructions could be embedded in the case law content.
Ingestion points: Data is fetched from the SAOS REST API via scripts/saos.py.
Boundary markers: The instructions do not define explicit delimiters to isolate the external judgment text from the agent's internal reasoning or commands.
Capability inventory: The skill uses the Bash tool to execute Python scripts and has access to Write for potentially storing results.
Sanitization: The Python script performs basic HTML tag removal for snippets but does not implement content filtering to detect or neutralize embedded prompt injection patterns.

saos-orzecznictwo