The Agent Skills Directory

[SAFE]: The skill implements a robust security auditing workflow, dividing the task into a reconnaissance phase (mapping attack surfaces) and an audit phase (deep analysis). This structure promotes thoroughness and minimizes the risk of missing vulnerabilities.
[SAFE]: The provided recon.py utility is a deterministic scanner that efficiently identifies security-sensitive code patterns. It correctly handles diverse character encodings (UTF-8, CP1251, Latin-1), which is essential for auditing legacy PHP environments like Bitrix without losing data during analysis.
[SAFE]: The instructions include strong guardrails against misinformation by requiring the agent to independently verify any findings provided in external context files (like pre-result.txt) against the actual module source code before including them in the final report.
[SAFE]: The skill's data handling is localized; it reads from specified module directories and saves reports to a local reports/ folder. No unauthorized network operations, data exfiltration attempts, or hardcoded credentials were found in the skill's instructions or scripts.
[SAFE]: The use of dynamic commands (e.g., date +%s) is limited to benign administrative tasks such as generating unique request IDs and does not involve the injection of untrusted user input into the shell environment.

scan