database-tactics
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides numerous commands for the agent to execute shell-based security tools including
redis-cli,netexec,impacket-mssqlclient,psql,mysql,mongosh, andhydra. These tools are used for reconnaissance, credential brute-forcing, and executing remote code on database targets. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It instructs the agent to query external database services and read their contents (e.g., searching for 'flag' strings, listing tables, and scanning keys). The instructions do not include boundary markers or data sanitization steps, which could allow an attacker to place malicious instructions inside a database that the agent might then follow when processing the results.
- Ingestion points: Database query results obtained through commands like
SELECT,KEYS *, andSCANacross multiple database types in SKILL.md and reference files. - Boundary markers: None present in the instructions to separate untrusted database output from agent instructions.
- Capability inventory: The agent has access to powerful shell commands and database clients capable of file I/O and further network activity across all scripts.
- Sanitization: No sanitization or validation of data retrieved from the databases is mentioned.
Audit Metadata