The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous curl and sqlmap command examples for security testing. These commands are intended to demonstrate how to store and trigger SQL injection payloads on a target system and do not involve executing untrusted remote code.\n- [EXTERNAL_DOWNLOADS]: The documentation includes references to attacker.burpcollaborator.net. This is a well-known service provided by PortSwigger for Out-of-Band (OOB) security testing and is standard for identifying blind vulnerabilities.\n- [INDIRECT_PROMPT_INJECTION]: The agent.py script processes external data sources which could contain malicious instructions.\n
Ingestion points: The scan_database_values function in scripts/agent.py reads JSON database dumps, and scan_source_code reads multiple source file formats.\n
Boundary markers: None present; the script relies on the provided file paths and assumes standard formats for analysis.\n
Capability inventory: The script's capabilities are restricted to file system read access and writing the final report to a local file. It does not have network access or dynamic code execution features.\n
Sanitization: Content is processed using standard JSON parsing and regular expressions without additional sanitization filters.

performing-second-order-sql-injection