The Agent Skills Directory

[SAFE] (SAFE): No technical vulnerabilities or malicious patterns were identified across all 10 threat categories. The skill is composed entirely of markdown documentation and guidance for the AI agent.\n- [NO_CODE] (SAFE): The skill package contains no Python scripts, Node.js packages, or shell commands, which effectively eliminates the surface for remote code execution, system persistence, or privilege escalation.\n- [PROMPT_INJECTION] (SAFE): The skill includes explicit instructions in 'SKILL.md' and 'EXAMPLES.md' to refuse requests for fabrication or unethical content generation, serving as a defensive layer against prompt injection attempts.\n- [DATA_EXFILTRATION] (SAFE): The intake process specifically identifies sensitive information types (e.g., security details, private roadmap) as off-limits for the skill, and the absence of network capabilities prevents data from being sent externally.\n- [INDIRECT_PROMPT_INJECTION] (SAFE): While the skill ingests user-provided data, it lacks any dangerous capabilities (exec/eval, network send, or file write) that would allow an injection to move beyond the immediate conversational context.

brand-storytelling