The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text provided by the user or other skills to verify legal citations. It lacks explicit boundary markers or sanitization instructions to prevent the agent from following malicious instructions hidden within the analyzed legal content.
Ingestion points: The skill takes "in input un testo" (a text as input) which is typically user-supplied or output from another skill.
Boundary markers: There are no instructions to use delimiters or ignore embedded instructions within the processed text.
Capability inventory: The skill utilizes WebFetch, browser automation tools, and file system write operations.
Sanitization: No sanitization, escaping, or filtering of the external content is described.
[COMMAND_EXECUTION]: The skill leverages high-privilege browser automation tools (mcp__Claude_in_Chrome) to navigate, interact with forms, and read console messages on the Italian Supreme Court's website (italgiure.giustizia.it). While this is the intended functionality, it represents an automated interaction with external web sessions.
[CREDENTIALS_UNSAFE]: The skill performs read and write operations on a state file located in the user's home directory (~/.claude/plugins/config/beccaria/state.json). This is used to track session-specific metadata like whether introductory messages have been displayed.
[EXTERNAL_DOWNLOADS]: The skill fetches information from multiple official legal databases via WebFetch, including Normattiva, EUR-Lex, and various Italian government authority portals. These are recognized official sources required for the skill's primary purpose.

verifica-fonti