The Agent Skills Directory

[PROMPT_INJECTION]: Documentation and rule definitions in SKILL.md and references/taxonomy.md contain example injection patterns. These are used for educational and detection purposes; the skill includes explicit meta-instructions to treat audited content as data, not commands.
[COMMAND_EXECUTION]: The fallback YARA engine in scripts/yara_lite.py uses eval() to process rule conditions from local signature files. This is a functional requirement for the rule engine and is not exposed to untrusted external input.
[DATA_EXFILTRATION]: Patterns matching exfiltration techniques in rules/malware.yar and scripts/analyzers.py are detection signatures used to identify threats in other skills.
[EXTERNAL_DOWNLOADS]: The skill documentation suggests cloning remote repositories for analysis, which is a necessary function for a security auditing tool and is initiated by explicit user request.

skill-security