The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts szukaj_orzeczen.py, saos_fetch.py, and raport_tematyczny.py utilize os.system() to invoke the Python package manager (pip) for installing the python-docx library if it is missing from the environment.
[DYNAMIC_EXECUTION]: The skill's workflow documentation (tryb-wyszukiwanie.md, raport-tematyczny.md) includes python3 -c commands to execute inline scripts for probing the API and performing statistical analysis on judgment data.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves legal judgment text from the SAOS API and instructs the agent to perform thematic analysis and text mining on this data. This introduces a surface where malicious content within court records could attempt to influence the agent's logic.
Ingestion points: Full judgment content (textContent) is retrieved from the SAOS API by saos_fetch.py and provided to the agent for analysis.
Boundary markers: While the skill includes instructions to check user-supplied search queries for prompt injection, it does not define specific delimiters or security markers for the judgment data fetched from the external API.
Capability inventory: The skill uses the Bash and Write tools to execute Python scripts, write reports to the file system, and download dependencies.
Sanitization: The scripts use strip_html() for content formatting, but there is no specialized sanitization to detect or neutralize prompt injection attempts within the legal texts.
[EXTERNAL_DOWNLOADS]: The skill fetches legal records from the official SAOS API (saos.org.pl) and downloads the python-docx library from the Python Package Index (PyPI).

szukaj-orzeczen-v2