The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's auditing process involves reading and analyzing external data sources which represents a potential surface for indirect prompt injection. * Ingestion points: The agent is instructed to read database schema metadata (table names, policy expressions) and a user-defined intent file (docs/rls-intent.md). * Boundary markers: The instructions do not specify the use of delimiters or 'ignore' markers when processing the database output. * Capability inventory: The skill leverages the agent's ability to execute database queries and read local documentation files. * Sanitization: There is no explicit requirement for the agent to sanitize the incoming metadata before performing the security diffing analysis.

rls-audit