The Agent Skills Directory

[SAFE]: The skill operates as a standard log parser without any malicious behavior. It reads input logs and writes reports to the local file system based on user-defined parameters.
[EXTERNAL_DOWNLOADS]: The skill mentions and imports the 'requests' library, which is a well-known and trusted package. No untrusted or suspicious external code is downloaded or executed.
[INDIRECT_PROMPT_INJECTION]: The script processes untrusted data (web application logs) which inherently contains attacker-controlled strings. This is expected behavior for a security analysis tool.
Ingestion points: Log file contents read by 'scripts/agent.py'.
Boundary markers: None; findings are exported as structured JSON objects.
Capability inventory: Local file read/write operations via 'argparse' arguments.
Sanitization: URI payloads are truncated to 500 characters in the output findings.

detecting-sql-injection-via-waf-logs