Load relevant skills for work sessions

Load and manage skills within GSD workflows with bounded learning guardrails

Ensure learned knowledge persists across sessions and subagent contexts without token bloat or unsafe refinements.

• · phase_name
• · skill_names
• · skill_set (project vs user level)

• · loaded_skills list
• · token_budget allocation
• · pending_suggestions list

• · .claude/commands/
• · ~/.claude/commands/

• · Skill files exist in correct directory
• · Token budget defined
• · Phase identity known

• · Stale conversation history consumes context
• · Skill loading pipeline fails silently on cache errors

• · 6-stage loading pipeline (Score, Resolve, ModelFilter, CacheOrder, Budget, Load)
• · 20% max content change per refinement
• · 7-day cooldown between refinements