token-efficiencylisted

# token-efficiency: Audit and optimize token consumption in Copilot CLI configs Apply evidence-based token-reduction techniques to Copilot CLI agent prompts, skill files, and hook injection scripts. Identify waste without sacrificing clarity or functionality. ## Scope This skill targets: - Agent prompt files (`copilot/agents/*.agent.md`) - Skill files (`copilot/skills/<x>/SKILL.md`) - Hook scripts that inject context into every turn or session (`copilot/hooks/scripts/*.sh`) - Config JSON files loaded at startup (not injected into model context) **Out of scope:** Code logic, test content, runtime behavior. Only text that enters the model's context window. ## Decision tree: Is this optimization safe? ### Safe to cut (high confidence) 1. **Redundant block repeats within a single file** — same prose/examples appearing twice in the same prompt. - Example: "JSON Output Format" section followed by a worked example with the same schema block. - Fix: keep one canonical version, point to it from the other location. - Risk: none if the pointer is clear. 2. **Verbose preamble/explanation** — conversational narrative that restates requirements already stated more concisely elsewhere. - Example: "This command generates feature-specific architecture... which merges into a pre-built HTML template. This two-stage approach..." (narrative) vs. "Generate feature architecture as structured JSON (injected into HTML template by a deterministic script)." (directive). - Fix: st