Diagnose

MANDATORY PREPARATION

Invoke /agent-workflow — it contains workflow principles, anti-patterns, and the Context Gathering Protocol. Follow the protocol before proceeding — if no workflow context exists yet, you MUST run /teach-maestro first.

Perform a systematic diagnostic scan across 5 dimensions. For each dimension, score 1-5 and provide specific findings.

Dimension 1: Prompt Quality (1-5)

Evaluate:

• Structure (4-zone pattern: role, context, instructions, output)
• Output schema definition (explicit vs. implicit)
• Instruction clarity (specific vs. vague)
• Edge case handling (addressed vs. ignored)
• Anti-patterns present (wall of text, contradictions, implicit format)

Dimension 2: Context Efficiency (1-5)

Evaluate:

• Context budget allocation (planned vs. ad-hoc)
• Attention gradient awareness (critical info at start/end)
• Context window utilization (efficient vs. wasteful)
• State management (explicit vs. implicit)
• Memory strategy (appropriate for conversation length)

Dimension 3: Tool Health (1-5)

Evaluate:

• Tool count (3-7 ideal, 13+ problematic)
• Description quality (specific vs. vague)
• Error handling (graceful vs. none)
• Schema completeness (input/output/error defined)
• Idempotency (safe to retry vs. side-effect prone)
• Scope attribution: Distinguish between project-configured tools (e.g., custom scripts, project MCP servers) and agent-level tools (e.g., built-in IDE tools, global MCP servers). Only flag tool overhead for tools the project can actually control

Dimension 4: Architecture Fitness (1-5)

Evaluate:

• Topology appropriateness (single vs. multi-agent justified)
• Agent boundaries (clear vs. overlapping)
• Handoff protocols (structured vs. ad-hoc)
• Observability (decisions logged vs. black box)
• Cost awareness (budgeted vs. unbounded)

Dimension 5: Safety & Reliability (1-5)

Evaluate:

• Input validation (present vs. absent)
• Output filtering (PII, content policy) — scope contextually: data flowing between a user's own frontend and backend (e.g., authenticated sessions, internal APIs) is lower risk than data exposed to external services or third-party APIs
• Cost controls (ceilings set vs. unbounded)
• Error recovery (fallbacks vs. crash)
• Evaluation strategy (golden tests vs. "it seems to work")

Diagnostic Report Format

╔══════════════════════════════════════╗
║          MAESTRO DIAGNOSTIC         ║
╠══════════════════════════════════════╣
║ Prompt Quality      ████░  4/5      ║
║ Context Efficiency   ███░░  3/5      ║
║ Tool Health          ██░░░  2/5      ║
║ Architecture         ████░  4/5      ║
║ Safety & Reliability ██░░░  2/5      ║
╠══════════════════════════════════════╣
║ Overall Score:       15/25           ║
╚══════════════════════════════════════╝

CRITICAL FINDINGS:
1. [Most severe issue — immediate action needed]
2. [Second most severe]
3. [Third]

RECOMMENDED ACTIONS:
1. Run /fortify to add error handling (addresses Tool Health + Safety)
2. Run /streamline to reduce tool count (addresses Tool Health)
3. Run /refine for prompt structure improvements (addresses Prompt Quality)

Maestro Command Mapping

Every recommended action MUST reference the specific Maestro command that addresses it. Use this mapping:

Do NOT give generic manual actions (e.g., "Add Vitest", "Create a rollback script") without also specifying which Maestro command the user should run to implement it. The recommended action format is:

Run /<command> to [specific action] (addresses [Dimension] #[gap number])

Scoring Guide

Diagnostic Checklist

• All 5 dimensions scored with specific evidence
• Critical findings listed in priority order
• Each finding includes specific file/component location
• Recommended actions reference specific Maestro commands (see Command Mapping above)
• Overall score calculated and report generated

Recommended Next Step

After diagnosis, run the command mapped to your lowest-scoring dimension. For a general improvement sequence: /fortify → /streamline → /refine.

NEVER:

• Give all 5s unless the workflow is genuinely production-excellent
• Skip dimensions — score all 5 even if some seem fine
• Diagnose without reading the actual workflow code/config
• Recommend changes without specific findings to support them
• Give generic manual actions without mapping them to a Maestro command