Agent skill · software engineering · jeremylongshore
schema-optimization-orchestrator
Multi-phase schema optimization workflow orchestrator. Creates session directories, spawns phase agents sequentially, validates outputs, aggregates results. Trigger: "run schema optimization", "optimize schema workflow", "execute schema phases
Why this skill is useful
Adds a multi-phase orchestration script for schema optimization that automates validation and evidence collection, which is not commonly available.
What it needs
Requires bash installed locally. About 3k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Schema Optimization Orchestrator Runs a multi-phase schema optimization workflow with strict validation and evidence collection. Workflow Pattern This is a test harness pattern: Creates isolated session directory per run Spawns 5 phase agents sequentially Each phase reads reference docs, runs scripts, writes reports Validates JSON outputs and file artifacts Aggregates final summary Inputs (JSON) Required: skilldir: Absolute path to this skill directory inputfolder: Path to data to analyze extractiontype: Type of data extraction (e.g., "bigqueryjson") Optional: sessiondirbase: Where to create run directories (default: reports/runs) Orchestration Steps 1. Create Session Directory 2. Run Phase 1: Initial Schema Analysis Spawn Phase 1 agent with: Expected output: Validation: JSON parse succeeds status is "complete" reportpath file exists schemasummary has required keys 3. Run Phase 2: Field Utilization Analysis Spawn Phase 2 agent with: Expected output: 4. Run Phase 3: Impact Assessment Spawn Phase 3 agent with: Expected output: 5. Run Phase 4: Verification with Script Spawn Phase 4 agent with: Expected output: 6. Run Phase 5: Final Recommendations Spawn Phase 5 agent with: Expected output: Output (JSON Only) Error Handling If any phase fails: Stop execution Return error status with phase details Preserve partial reports for debugging Validation Rules After each phase: 1. Parse returned JSON (fail if invalid) 2. Check status is "complete" (fail if not) 3. Verify reportpath exists on disk (fail if not) 4. Validate phase-specific summary keys (fail if missing) Implementation Notes Use Task tool to spawn phase agents Pass exact file paths (no wildcards) Session directory must be absolute path All reports must be written before returning No terminal output except final JSON Example Usage Files Created Per Run Each file is evidence of work completed.
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/schema-optimization