The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to audit database configuration files and system command outputs, which are external data sources that could potentially contain malicious instructions. However, the risk is mitigated by the skill's focus on specific security patterns and read-only inspection methods.
Ingestion points: The agent is instructed to read local configuration files (e.g., pg_hba.conf, postgresql.conf, my.cnf, redis.conf, mongod.conf) and process the output of system commands like ss, psql, mysql, redis-cli, and mongosh.
Boundary markers: The skill provides structured tables and specific mental models to guide the agent in categorizing findings, which helps maintain clear boundaries between the data being inspected and the agent's instructions.
Capability inventory: The skill uses standard shell utilities (grep, awk, find, ls) and official database CLI tools to perform its tasks. It does not include automated remediation that would execute changes without review.
Sanitization: The audit process relies on extracting specific, well-defined security configuration keys and SQL results, reducing the likelihood of executing arbitrary text embedded in those files.

aegis-db-server-hardening