This updates token/cost optimization guidance to explicitly treat sub-skills as progressive disclosure. The main prompt is now positioned as a compact execution plan, with detailed report-format and rubric instructions deferred to on-demand sub-skills to reduce ambient context and cost.

• Token optimization guide updates (.github/aw/token-optimization.md)

  • Added a checklist item for Sub-skills alongside existing cost levers.
  • Added explicit guidance to:
    • keep the main prompt plan-oriented,
    • move verbose formatting/playbook content into ## skill: blocks,
    • invoke skills only at the phase where they are needed (e.g., final output).

• Cost management reference updates (docs/src/content/docs/reference/cost-management.md)

  • Expanded inline-skills section with a clear split:
    • main prompt = sequencing + decisions,
    • sub-skills = detailed templates/rubrics.
  • Added explicit “invoke late, not at startup” framing to reinforce lower token overhead.

• Illustrative pattern

  Triage the issue and decide next action.
  
  ## skill: `report-layout`
  ---
  description: Final report format and section requirements.
  ---
  Use this only when producing the final report body.