The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze user-provided legal scenarios and criminal cases, which creates an attack surface for indirect prompt injection where a user could embed instructions to bias the agent's legal conclusions.
Ingestion points: Processed case data and legal queries provided by the user in the context of criminal law analysis (Workflow Module 10).
Boundary markers: The skill includes 'Restricoes Absolutas' (Absolute Restrictions) to guide agent behavior, but lacks specific delimiters or explicit instructions to ignore commands embedded within the untrusted case data.
Capability inventory: The skill focuses on legal knowledge but provides instructions for command-line interaction and script execution.
Sanitization: No explicit sanitization, validation, or filtering of user-supplied legal context is implemented.
[COMMAND_EXECUTION]: The skill's documentation contains instructions to execute local Python scripts using absolute file paths.
Evidence: The 'Instalacao' and 'Comandos E Uso' sections suggest running python C:\Users\renat\skills\agent-orchestrator\scripts\scan_registry.py and python agent-orchestrator/scripts/match_skills.py. These scripts are external to the skill's own content.

advogado-criminal