Concurrent Skill Execution & Real-Time Iterative Improvement

# Protocol: Concurrent Skill Execution & Real-Time Iterative Improvement ## Mission You are tasked with executing a specific Agent Skill on a real user task, while simultaneously acting as a meta-auditor to evaluate and upgrade the skill's instruction set (`SKILL.md`). --- ## Two-Track Execution Workflow: ### Track A: Production Task Execution 1. Load and strictly follow the skill's existing `SKILL.md` instructions. 2. Execute all required research, verification, and output generation steps without skipping quality gates. 3. Deliver the primary task output with high precision. ### Track B: Skill Meta-Audit & Optimization As you execute Track A, log any friction points across these 4 categories: 1. **Ambiguous Instructions:** Where did you need to make assumptions because the rules were vague? 2. **Missing Edge Cases:** What real-world conditions did the existing skill instructions fail to anticipate? 3. **Inefficient Steps:** Which actions generated token bloat or redundant tool calls? 4. **Verification Gaps:** What quality checks should be made mandatory before task completion? --- ## Final Output Structure: 1. **Primary Task Deliverable:** Complete, high-quality result requested by the user. 2. **Skill Improvement Report:** - Identified friction points with line-level references to `SKILL.md`. - Proposed diff / drop-in replacement for the improved `SKILL.md` sections.