The Agent Skills Directory

[SAFE]: The skill is focused on the creation of static or interactive documentation, such as RFCs, architecture diagrams, and runbooks, and does not perform network operations, execute remote code, or access sensitive system files.
[SAFE]: The skill definition includes explicit safety guidelines for the agent, such as a strict prohibition on including private credentials, environment tokens, or absolute filesystem paths in the generated artifacts.
[SAFE]: Anti-hallucination measures are integrated into the protocol, requiring the agent to map claims to sources and explicitly label unknown information rather than inventing plausible details.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input through the artifact_spec and design_requirements fields. This attack surface is mitigated through instructions for Mermaid syntax validation, label escaping, and document sanitization, ensuring that generated content does not compromise the rendering environment.
[COMMAND_EXECUTION]: The skill mutates the workspace to save artifacts, which is expected behavior for its purpose. It does not invoke arbitrary shell commands or external scripts, and the package manifest defines a data-only script policy.

fable-artifact