The Agent Skills Directory

[SAFE]: The skill implements a robust multi-agent orchestration pattern for text processing (detecting AI patterns, injecting voice, and compressing text). All operations are local to the provided text context.
[SAFE]: The skill uses allowed tools such as Read, Grep, Glob, Bash, WebSearch, and WebFetch within the documented scope of analyzing content and fetching brand/SEO guidelines.
[SAFE]: The 'Absolute Prohibitions' and '47-pattern checklist' are internal logic constraints for the LLM to improve content quality (e.g., forbidding em dashes or specific AI-typical phrases) and do not represent malicious behavior.
[SAFE]: The bootstrap-experience.ts script is a utility for managing local project metadata (.agents/experience/) to avoid redundant user questioning. It includes safety checks to prevent writing through symbolic links and ensures it operates within the project directory.
[SAFE]: There is no evidence of prompt injection, data exfiltration, or unauthorized privilege escalation. External references point to standard internal project files (e.g., brand/BRAND.md, research/product-context.md).

humanize