ai-act-triage-pl
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses three local Python scripts (
klasyfikator_ryzyka_ai.py,plan_zgodnosci.py,tracker_obowiazkow.py) to process compliance data. These scripts rely solely on the Python standard library and do not perform network operations or access sensitive file paths. - [INDIRECT_PROMPT_INJECTION]: The skill processes external information about AI systems to provide a legal triage. While this constitutes an ingestion surface, the risk is mitigated by the skill's design, which explicitly requires human approval for all outputs and includes clear warnings that the results are drafts for legal review, not final decisions. The scripts use deterministic logic (decision trees) rather than LLM-based reasoning for the classification steps.
- [SAFE]: The skill follows security best practices by utilizing local execution (
data-residency: local), requiring human approval (requires-human-approval: true), and providing clear attribution for its legal references. The tools listed inallowed-tools(Read, Bash) are scoped to the skill's directory for reading documentation and running the included analysis scripts.
Audit Metadata