The Agent Skills Directory

[SAFE]: The skill provides a task-orchestration framework and does not contain any malicious commands, obfuscation, or data exfiltration patterns. It focuses on process management and agentic workflows.
[PROMPT_INJECTION]: The workflow involves research subagents that fetch documentation or search the web. While this introduces an indirect prompt injection surface from untrusted external data, the skill mitigates this risk by including a mandatory 'adversarial verification' stage where separate subagents attempt to refute and break the work before completion.
[EXTERNAL_DOWNLOADS]: The skill references a sibling skill ('fable-method') expected to be present in local directories. This is a standard dependency for modular skills and does not involve remote code execution.

fable-loop