The Agent Skills Directory

[SAFE]: The file SKILL.md describes a strict multi-phase human-in-the-loop workflow for product managers or product owners to refine requirements, draft a PRD, generate structured development tasks, create ticket definitions, and lay out sprint boundaries. All operations strictly require explicit user approvals ('hard gates') before stepping forward, ensuring that the AI agent cannot take unilateral actions or bypass user instructions.
[SAFE]: Orchestration uses internal, atomic sub-skills (create-prd, generate-tasks, plan-tickets) which write outputs locally to /tasks/ in standard Markdown format. No untrusted third-party packages, remote scripts (curl | bash), or network calls are utilized.
[SAFE]: No obfuscation, payload injection, dynamic context execution triggers, metadata poisoning, or credential exposure vectors exist within the asset.

product-owner