The Agent Skills Directory

[SAFE]: No security issues detected. The skill follows best practices for AI agent safety and defensive security auditing.
[INDIRECT_PROMPT_INJECTION]: The skill identifies and mitigates risks associated with processing untrusted data. It provides specific instructions for agents to ignore PR-controlled configuration files (e.g., AGENTS.md, .claude/) and use read-only environments when auditing external code.
[DATA_EXFILTRATION]: The skill performs research by fanning out requests to well-known security domains such as NIST, CISA, and OWASP. These network operations are strictly for information gathering from trusted sources and align with the skill's purpose.
[COMMAND_EXECUTION]: The instructions provide safe code patterns for parameterizing SQL queries and avoiding shell execution to prevent common injection attacks. The skill does not execute any commands on the host system.

security-checklist