@dzhechkov/skills-bto 1.0.0

package/templates/.claude/skills/bto/examples/sample-eval-report.md

@@ -0,0 +1,436 @@
+# Sample BTO Evaluation Report
+
+> This file is a **reference example** showing the complete output format of a BTO TEST evaluation.
+> Use it as the authoritative template when formatting evaluation reports.
+>
+> **Hypothetical subject:** `.claude/skills/data-pipeline-validator/` — a skill for validating ETL pipeline configurations.
+
+---
+
+## Report Header
+
+```
+╔══════════════════════════════════════════════════════════════╗
+║           BTO EVALUATION REPORT — TEST MODULE               ║
+╠══════════════════════════════════════════════════════════════╣
+║  Artifact:      data-pipeline-validator (Skill)             ║
+║  Path:          .claude/skills/data-pipeline-validator/     ║
+║  Evaluated:     2026-03-01 14:32 UTC                        ║
+║  Evaluator:     BTO TEST v1.2                               ║
+║  Layers run:    Layer 0 + Layer 1 + Layer 2                 ║
+╚══════════════════════════════════════════════════════════════╝
+```
+
+---
+
+## Layer 0: Deterministic Pre-Checks
+
+**Artifact type:** Skill
+**Applicable checks:** Universal (12) + Skill-specific (16) = 28 total
+
+### Universal Checks (12 applicable)
+
+| ID | Check | Result | Note |
+|----|-------|--------|------|
+| U-01 | File exists and is non-empty | PASS | SKILL.md: 8.4 KB |
+| U-02 | UTF-8 encoding | PASS | |
+| U-03 | Starts with level-1 heading | PASS | `# Data Pipeline Validator` |
+| U-04 | No placeholder text | FAIL | Found `[INSERT SCHEMA PATH HERE]` in modules/validate.md |
+| U-05 | No empty sections | PASS | |
+| U-06 | Consistent heading hierarchy | PASS | |
+| U-07 | No broken internal cross-references | FAIL | `references/schema-types.md` referenced but missing |
+| U-08 | File size within bounds | PASS | 8.4 KB (within 200B–100KB) |
+| U-09 | No trailing whitespace | PASS | Auto-fixed during generation |
+| U-10 | Standard Markdown only | PASS | |
+| U-11 | Code blocks properly closed | PASS | |
+| U-12 | No duplicate top-level sections | PASS | |
+
+**Universal result: 10 / 12**
+
+### Skill-Specific Checks (16 applicable)
+
+| ID | Check | Result | Note |
+|----|-------|--------|------|
+| SK-01 | SKILL.md exists at skill root | PASS | |
+| SK-02 | Has `## Overview` section | PASS | |
+| SK-03 | Has `## Anti-Patterns` section | PASS | |
+| SK-04 | Has Quick Start or usage example | PASS | Section present with 2 examples |
+| SK-05 | `modules/` directory exists | PASS | |
+| SK-06 | All modules referenced exist on disk | PASS | 3 modules, all present |
+| SK-07 | `references/` directory exists | PASS | |
+| SK-08 | All references referenced exist on disk | FAIL | `schema-types.md` missing (linked at line 47) |
+| SK-09 | `examples/` directory exists | PASS | |
+| SK-10 | At least one example file present | FAIL | Directory exists but is empty |
+| SK-11 | Skill name matches directory name | PASS | |
+| SK-12 | Has `## Dependencies` section | PASS | |
+| SK-13 | SKILL.md size within skill bounds | PASS | 8.4 KB |
+| SK-14 | Total directory size within bounds | PASS | 22 KB total |
+| SK-15 | Each module has `# Title` heading | PASS | |
+| SK-16 | No circular cross-references | PASS | |
+
+**Skill-specific result: 13 / 16**
+
+### Layer 0 Summary
+
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+LAYER 0 PRE-FLIGHT CHECK
+Artifact: data-pipeline-validator (Skill)
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+Universal checks:       10 / 12
+Skill-specific checks:  13 / 16
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+TOTAL:                  23 / 28
+Pass rate:              82.1%
+Status:                 PASS (threshold: 80%)
+
+Failed checks:
+  - [U-04]  Placeholder text in modules/validate.md line 23 *
+  - [U-07]  Missing file: references/schema-types.md
+  - [SK-08] Broken reference: schema-types.md (line 47 in SKILL.md)
+  - [SK-10] Empty examples/ directory
+
+(* = auto-fixable)
+
+Verdict: PROCEED TO LAYER 1 (with noted issues logged)
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+**Decision:** Layer 0 PASS (82.1% > 80%). Flagged issues logged — will be reflected in Layer 1/2 scoring under CORRECTNESS and USABILITY dimensions.
+
+---
+
+## Layer 1: Single LLM Judge (Quick Mode)
+
+**Model:** claude-haiku (cost-optimized)
+**Purpose:** Rapid spot-check before committing to full 3-judge panel
+
+**Judge Prompt Used:**
+```
+You are a Claude Code artifact evaluator. Rate the skill at the provided path
+on 5 dimensions (1-10 each). Provide a score, one-sentence justification,
+and top 3 improvement suggestions. Artifact type: Skill.
+```
+
+### Layer 1 Scores
+
+| Dimension | Score | Justification |
+|-----------|-------|---------------|
+| CLARITY | 7 | Instructions are mostly clear but the validation protocol section has ambiguous branching logic at step 4 |
+| COMPLETENESS | 6 | Missing example file and one broken reference significantly reduce perceived completeness |
+| ACTIONABILITY | 7 | Core happy path is followable; failure paths are underspecified |
+| QUALITY | 7 | Well-structured with good use of tables; prose in modules could be tighter |
+| ANTI-PATTERNS | 8 | Anti-patterns section is thorough with 6 patterns, each with signals and fixes |
+
+**Layer 1 Average: 7.0**
+**Threshold: 7.0**
+**Status: BORDERLINE PASS — proceed to Layer 2 for definitive evaluation**
+
+**Layer 1 Top 3 Suggestions:**
+1. Add at least one complete worked example showing a real ETL schema being validated end-to-end
+2. Fix the broken `schema-types.md` reference or remove the link
+3. Clarify the branching logic in step 4 of the validation protocol (what happens if schema is partially valid?)
+
+---
+
+## Layer 2: Full Judge Panel (3 Agents)
+
+**Models:** claude-sonnet (all 3 judges)
+**Rubric source:** `references/judge-rubrics.md`
+**Execution:** Parallel (all 3 judges ran simultaneously)
+**Aggregation:** Weighted average — Expert 0.4 / Critic 0.3 / Auditor 0.3
+
+---
+
+### Judge 1: Domain Expert
+
+**Role:** Evaluates accuracy, depth, methodology, and domain fit.
+**Calibration:** Balanced — scores represent genuine quality assessment.
+
+| Dimension | Score | Expert Commentary |
+|-----------|-------|-------------------|
+| METHODOLOGY | 8 | The validation protocol follows a logical sequence: schema → types → constraints → relationships. The hierarchical validation approach is sound for ETL contexts. |
+| DEPTH | 6 | Covers the primary validation scenarios (schema drift, type mismatches, null handling) but misses temporal consistency checks and incremental load edge cases — critical for production pipelines. |
+| CORRECTNESS | 7 | All referenced validation patterns are real and applicable. The one broken reference (`schema-types.md`) is a quality concern but doesn't indicate incorrect claims. |
+| USABILITY | 7 | Quick Start is useful. Navigation between SKILL.md and modules is smooth. The empty examples directory is a significant usability gap — users need a worked example. |
+| ROBUSTNESS | 8 | Anti-patterns section is strong. Handles schema evolution, backwards compatibility, and partial validation failure. Missing: network timeout handling for remote schema fetches. |
+
+**Expert subtotal: (8+6+7+7+8)/5 = 7.2**
+
+---
+
+### Judge 2: Critic
+
+**Role:** Identifies gaps, weaknesses, anti-patterns, and edge cases.
+**Calibration:** Strict — intentionally scores conservative to surface issues.
+
+| Dimension | Score | Critic Commentary |
+|-----------|-------|-------------------|
+| METHODOLOGY | 6 | The protocol works for simple schemas but collapses under nested/recursive schema structures. Step 4 (constraint validation) assumes a flat schema model. No guidance on when to abort vs. continue validation after first failure. |
+| DEPTH | 5 | No coverage of: streaming data validation, schema registry integration (Confluent, AWS Glue), or multi-source join validation. The "data types" coverage only addresses SQL types — no Avro, Parquet, or Arrow type systems. |
+| CORRECTNESS | 6 | The placeholder `[INSERT SCHEMA PATH HERE]` in modules/validate.md (line 23) is a clear correctness failure — the skill was shipped incomplete. The claim about "99.9% detection rate" on line 31 has no source. |
+| USABILITY | 5 | Zero examples is a dealbreaker for a validation skill. Users cannot learn by example. The broken reference creates a dead end. Anti-patterns section, while good, is buried at the end — should appear earlier as a pre-check. |
+| ROBUSTNESS | 6 | Handles the happy path and common failures. No coverage of: what to do if the target system is unavailable, concurrent validation conflicts, or validation against evolving schemas mid-run. |
+
+**Critic subtotal: (6+5+6+5+6)/5 = 5.6**
+
+---
+
+### Judge 3: Completeness Auditor
+
+**Role:** Checks structure, coverage, cross-references, and actionability.
+**Calibration:** Neutral — focuses on structural completeness, not quality of content.
+
+| Dimension | Score | Auditor Commentary |
+|-----------|-------|-------------------|
+| METHODOLOGY | 7 | All required sections present in SKILL.md. Module structure follows conventions. Decision flow between modules is documented. Protocol steps are numbered and sequential. |
+| DEPTH | 6 | 3 modules present: validate.md, report.md, fix-suggestions.md — reasonable coverage. However, no module for schema-registry integration or configuration management, which the skill description implies are supported. |
+| CORRECTNESS | 6 | 2 structural correctness failures found: (1) broken reference to schema-types.md, (2) placeholder text in validate.md. These are Layer 0 failures that reached Layer 2 — the skill was not fully completed before evaluation. |
+| USABILITY | 7 | Navigation is clear. Table of contents in SKILL.md maps correctly to modules. Quick Start section provides 2 invocation examples. Dependencies section complete. Loses points for missing examples. |
+| ROBUSTNESS | 7 | Anti-patterns section documents 6 patterns. Failure modes section in validate.md covers 4 scenarios. Abort conditions defined. Minor gap: no documentation of what constitutes a "critical" vs. "warning" validation finding. |
+
+**Auditor subtotal: (7+6+6+7+7)/5 = 6.6**
+
+---
+
+### Aggregated Scores (Weighted)
+
+**Formula:** Expert (×0.4) + Critic (×0.3) + Auditor (×0.3)
+
+| Dimension | Expert | Critic | Auditor | Weighted Score | Bar |
+|-----------|--------|--------|---------|----------------|-----|
+| METHODOLOGY | 8 | 6 | 7 | 7.1 | `███████░░░` |
+| DEPTH | 6 | 5 | 6 | 5.7 | `█████▒░░░░` |
+| CORRECTNESS | 7 | 6 | 6 | 6.4 | `██████░░░░` |
+| USABILITY | 7 | 5 | 7 | 6.4 | `██████░░░░` |
+| ROBUSTNESS | 8 | 6 | 7 | 7.1 | `███████░░░` |
+| **OVERALL** | **7.2** | **5.6** | **6.6** | **6.5** | `██████▒░░░` |
+
+**Bar legend:** `█` = full point, `▒` = half point, `░` = empty
+
+**Overall weighted score: 6.5 / 10**
+**Pass threshold: 7.0**
+**Status: BELOW THRESHOLD**
+
+---
+
+### Disagreement Analysis
+
+> Flag trigger: any dimension where `max - min > 3` across judges.
+
+| Dimension | Max | Min | Range | Disagreement Flag |
+|-----------|-----|-----|-------|-------------------|
+| METHODOLOGY | 8 | 6 | 2 | None |
+| DEPTH | 6 | 5 | 1 | None |
+| CORRECTNESS | 7 | 6 | 1 | None |
+| USABILITY | 7 | 5 | **2** | None (under threshold of 3) |
+| ROBUSTNESS | 8 | 6 | 2 | None |
+
+**No disagreement flags triggered.** Maximum range across all dimensions was 2 (USABILITY), below the trigger threshold of 3. Meta-judge not required.
+
+**Interpretation:** The judges are aligned in their assessment. The Critic's USABILITY score of 5 reflects the empty examples directory — a structural deficiency all judges identified, but weighed differently. The Expert and Auditor judged the existing structure more charitably; the Critic penalized the missing examples more severely. This spread is within normal calibration variance.
+
+---
+
+### Meta-Judge Status
+
+**Triggered:** NO
+**Reason:** All dimension ranges within tolerance (max range: 2, threshold: 3)
+**Action:** Proceed with weighted aggregation as final score.
+
+---
+
+## Top Improvements
+
+Priority ordered by weighted impact on overall score.
+
+### Priority 1 — Add a worked example (Impact: HIGH)
+**Failing checks:** SK-10, Layer 1 suggestion 1, Critic USABILITY: 5, Auditor USABILITY: 7
+**What to fix:** Create `examples/validate-postgres-to-s3.md` showing a complete validation run of a realistic PostgreSQL-to-S3 ETL pipeline: input schema, validation output, error report, and fix suggestions.
+**Expected score improvement:** USABILITY +1.5 to +2.0 points
+
+### Priority 2 — Create missing `references/schema-types.md` (Impact: HIGH)
+**Failing checks:** U-07, SK-08, Critic CORRECTNESS: 6
+**What to fix:** Create the referenced file with a type mapping table covering SQL, Avro, Parquet, and Arrow type systems. Remove the broken link if out of scope.
+**Expected score improvement:** CORRECTNESS +0.8, DEPTH +0.5
+
+### Priority 3 — Remove placeholder text (Impact: MEDIUM)
+**Failing checks:** U-04, Critic CORRECTNESS: 6
+**What to fix:** Replace `[INSERT SCHEMA PATH HERE]` in `modules/validate.md` line 23 with a concrete example path or a proper `$ARGUMENTS` reference.
+**Expected score improvement:** CORRECTNESS +0.3
+
+### Priority 4 — Expand depth for non-SQL type systems (Impact: MEDIUM)
+**Failing checks:** Critic DEPTH: 5, Expert DEPTH: 6
+**What to fix:** Add a module section or reference covering Avro, Parquet, and Arrow schemas. Add a note on schema registry integration (AWS Glue Data Catalog, Confluent Schema Registry).
+**Expected score improvement:** DEPTH +1.0 to +1.5
+
+### Priority 5 — Clarify critical vs. warning severity levels (Impact: LOW)
+**Failing checks:** Auditor ROBUSTNESS partial
+**What to fix:** Add a severity matrix to the validation output format — define which validation failures are CRITICAL (block pipeline) vs. WARNING (log and continue).
+**Expected score improvement:** ROBUSTNESS +0.3
+
+---
+
+## Overall Verdict
+
+```
+╔══════════════════════════════════════════════════════════════╗
+║                    FINAL VERDICT                            ║
+╠══════════════════════════════════════════════════════════════╣
+║                                                              ║
+║   Artifact:    data-pipeline-validator (Skill)              ║
+║   Score:       6.5 / 10                                     ║
+║                                                              ║
+║   Score breakdown:                                           ║
+║   METHODOLOGY  ███████░░░  7.1                              ║
+║   DEPTH        █████▒░░░░  5.7                              ║
+║   CORRECTNESS  ██████░░░░  6.4                              ║
+║   USABILITY    ██████░░░░  6.4                              ║
+║   ROBUSTNESS   ███████░░░  7.1                              ║
+║                ──────────                                    ║
+║   OVERALL      ██████▒░░░  6.5                              ║
+║                                                              ║
+║   Verdict:     NEEDS IMPROVEMENT                            ║
+║   Threshold:   7.0 (not met)                                ║
+║                                                              ║
+║   Recommended action:                                        ║
+║   → Apply Priority 1-3 fixes (est. 30 min)                 ║
+║   → Re-run /bto-test to verify improvement                  ║
+║   → Do NOT run /bto-optimize until score ≥ 7.0             ║
+║                                                              ║
+║   Estimated score after Priority 1-3 fixes:  7.5 – 8.0     ║
+╚══════════════════════════════════════════════════════════════╝
+```
+
+---
+
+## Appendix: Judge Prompts Used
+
+### Judge 1 (Domain Expert) Prompt
+
+```
+You are a Domain Expert evaluating a Claude Code skill artifact.
+Your role: assess accuracy, depth, methodology, and domain fit.
+Score each dimension 1-10 using the rubric in judge-rubrics.md.
+Be balanced — reward genuine quality, penalize real gaps.
+
+Artifact type: Skill
+Artifact path: .claude/skills/data-pipeline-validator/
+
+Dimensions to score:
+1. METHODOLOGY — Is the approach sound and well-structured?
+2. DEPTH — Is the content thorough enough for production use?
+3. CORRECTNESS — Are claims accurate and instructions valid?
+4. USABILITY — Can a practitioner effectively use this artifact?
+5. ROBUSTNESS — Does it handle edge cases and failure modes?
+
+Output: JSON with keys: methodology, depth, correctness, usability,
+robustness, commentary_per_dimension, top_3_issues
+```
+
+### Judge 2 (Critic) Prompt
+
+```
+You are a strict Critic evaluating a Claude Code skill artifact.
+Your role: find gaps, weaknesses, anti-patterns, and missing edge cases.
+Score calibration: your scores should average around 5-6 (strict).
+Do not reward potential — only reward what is actually present.
+
+Artifact type: Skill
+Artifact path: .claude/skills/data-pipeline-validator/
+
+Dimensions to score:
+1. METHODOLOGY — Where does the approach break down?
+2. DEPTH — What important topics are missing?
+3. CORRECTNESS — What is wrong, incomplete, or unverified?
+4. USABILITY — Where will users get stuck?
+5. ROBUSTNESS — What failure modes are unhandled?
+
+Output: JSON with keys: methodology, depth, correctness, usability,
+robustness, commentary_per_dimension, top_3_issues
+```
+
+### Judge 3 (Completeness Auditor) Prompt
+
+```
+You are a Completeness Auditor evaluating a Claude Code skill artifact.
+Your role: check structural completeness, cross-reference integrity,
+section coverage, and actionability.
+Be neutral — focus on what is and is not present, not on quality of content.
+
+Artifact type: Skill
+Artifact path: .claude/skills/data-pipeline-validator/
+
+Dimensions to score:
+1. METHODOLOGY — Are all required protocol elements present?
+2. DEPTH — Does the number and depth of sections match the scope?
+3. CORRECTNESS — Are all cross-references valid and content accurate?
+4. USABILITY — Is navigation clear and information findable?
+5. ROBUSTNESS — Are failure modes and edge cases documented?
+
+Output: JSON with keys: methodology, depth, correctness, usability,
+robustness, commentary_per_dimension, top_3_issues
+```
+
+---
+
+## Appendix: Raw Judge Output (JSON)
+
+```json
+{
+  "evaluation_metadata": {
+    "artifact": "data-pipeline-validator",
+    "type": "skill",
+    "timestamp": "2026-03-01T14:32:00Z",
+    "bto_version": "1.2"
+  },
+  "layer_0": {
+    "total": 28,
+    "passed": 23,
+    "pass_rate": 0.821,
+    "status": "PASS",
+    "failed_ids": ["U-04", "U-07", "SK-08", "SK-10"]
+  },
+  "layer_1": {
+    "model": "claude-haiku",
+    "scores": {
+      "clarity": 7, "completeness": 6, "actionability": 7,
+      "quality": 7, "anti_patterns": 8
+    },
+    "average": 7.0,
+    "status": "BORDERLINE_PASS"
+  },
+  "layer_2": {
+    "judges": {
+      "expert": {
+        "model": "claude-sonnet",
+        "scores": { "methodology": 8, "depth": 6, "correctness": 7, "usability": 7, "robustness": 8 },
+        "subtotal": 7.2
+      },
+      "critic": {
+        "model": "claude-sonnet",
+        "scores": { "methodology": 6, "depth": 5, "correctness": 6, "usability": 5, "robustness": 6 },
+        "subtotal": 5.6
+      },
+      "auditor": {
+        "model": "claude-sonnet",
+        "scores": { "methodology": 7, "depth": 6, "correctness": 6, "usability": 7, "robustness": 7 },
+        "subtotal": 6.6
+      }
+    },
+    "weighted": {
+      "methodology": 7.1, "depth": 5.7, "correctness": 6.4,
+      "usability": 6.4, "robustness": 7.1, "overall": 6.5
+    },
+    "disagreement_flags": [],
+    "meta_judge_triggered": false,
+    "status": "BELOW_THRESHOLD"
+  },
+  "verdict": {
+    "score": 6.5,
+    "threshold": 7.0,
+    "result": "NEEDS_IMPROVEMENT",
+    "recommended_action": "apply_priority_fixes_then_retest"
+  }
+}
+```

package/templates/.claude/skills/bto/modules/build.md

@@ -0,0 +1,189 @@
+# BUILD Module — Artifact Generation Protocol
+
+## Purpose
+
+Generate production-quality Claude Code artifacts (skills, commands, rules, agent templates) from natural language requirements.
+
+## Input
+
+- **Description:** Natural language description of what the artifact should do
+- **Type:** skill | command | rule | agent (auto-detected or explicit)
+- **Mode:** QUICK | DEEP (default: QUICK)
+- **References:** Optional paths to existing artifacts as examples
+
+## Protocol
+
+### Step 1: Type Detection
+
+If type is not explicitly specified, detect from description:
+
+| Signal | Detected Type |
+|--------|--------------|
+| "skill", "module", "capability", "protocol" | skill |
+| "command", "slash command", "/something", "pipeline" | command |
+| "rule", "constraint", "convention", "anti-pattern" | rule |
+| "agent", "worker", "parallel", "swarm" | agent |
+
+### Step 2: Requirements Gathering
+
+**QUICK mode** — Extract from description directly:
+1. Parse artifact name (kebab-case)
+2. Extract key capabilities
+3. Identify domain constraints
+4. Proceed to generation
+
+**DEEP mode** — Use `explore` skill:
+1. Read `.claude/skills/explore/SKILL.md`
+2. Follow explore protocol to clarify:
+   - Exact scope and boundaries
+   - Target users/consumers
+   - Input/output format
+   - Quality criteria
+   - Edge cases
+3. Produce requirements brief
+4. Confirm with user before generation
+
+### Step 3: Generation Templates
+
+#### Skill Template
+
+```
+.claude/skills/<name>/
+├── SKILL.md
+│   ├── # <Name>
+│   ├── ## Overview
+│   ├── ## Quick Start
+│   ├── ## Protocol
+│   │   ├── Step 1: ...
+│   │   ├── Step 2: ...
+│   │   └── Step N: ...
+│   ├── ## Output Format
+│   ├── ## Anti-Patterns
+│   └── ## Dependencies
+├── modules/          (if multi-module)
+│   └── <module>.md
+├── references/       (always include at least one)
+│   └── <ref>.md
+└── examples/         (always include at least one)
+    └── <example>.md
+```
+
+#### Command Template
+
+```markdown
+# /command-name — Short Description
+
+## Usage
+/command-name [arguments]
+
+## Parameters
+- $ARGUMENTS — Description
+
+## Protocol
+
+### Step 1: Setup
+- Validate arguments
+- Load required skills: Read `.claude/skills/<skill>/SKILL.md`
+
+### Step 2: Execution
+- Main logic here
+- Use Agent tool for parallelism where applicable
+
+### Step 3: Output
+- Create artifact files
+- Display checkpoint
+
+## Checkpoint
+═══════════════════════════════════════════════════════
+⏸️ CHECKPOINT: [Command Name] Complete
+...
+═══════════════════════════════════════════════════════
+```
+
+#### Rule Template
+
+```markdown
+# Rule Name
+
+## Patterns
+
+| Pattern | Detection Signal | Required Fix |
+|---------|-----------------|-------------|
+| ... | ... | ... |
+
+## Auto-Detection
+When generating content, self-check against these patterns.
+If detected, flag with ⚠️ and fix before proceeding.
+```
+
+#### Agent Template
+
+```markdown
+# Agent Name
+
+## Purpose
+What this agent does.
+
+## Configuration
+- Model: haiku | sonnet | opus
+- Isolation: reads X, writes Y
+- Max turns: N
+
+## Prompt Template
+```
+
+### Step 4: Self-Review
+
+After generation, validate against quality checklist:
+
+1. **Structure check:**
+   - Required sections present for artifact type
+   - No empty sections
+   - Proper markdown formatting
+
+2. **Content check:**
+   - No generic/placeholder content
+   - Specific to the domain
+   - Anti-patterns section populated
+   - At least one concrete example
+
+3. **Convention check:**
+   - File naming: kebab-case
+   - Directory naming: kebab-case
+   - Heading hierarchy: proper nesting
+   - References resolve to actual files
+
+4. **Size check:**
+   - SKILL.md: 2KB-30KB (ideal: 5KB-15KB)
+   - Module: 1KB-15KB
+   - Reference: 500B-10KB
+   - Example: 500B-5KB
+
+### Step 5: Output
+
+1. Create all files in the target directory
+2. Display summary:
+   ```
+   ✅ BUILD Complete
+   Artifact: .claude/skills/<name>/
+   Files created:
+     - SKILL.md (X KB)
+     - modules/<m>.md (X KB)
+     - references/<r>.md (X KB)
+     - examples/<e>.md (X KB)
+
+   Next: Run /bto-test .claude/skills/<name>/ to evaluate
+   ```
+
+## Anti-Patterns
+
+| Anti-Pattern | Detection | Fix |
+|-------------|-----------|-----|
+| Generic skill | No domain-specific terms | Add domain context and constraints |
+| Missing references | references/ empty | Add at least one reference file |
+| No examples | examples/ empty | Add at least one few-shot example |
+| Over-scoped | SKILL.md > 30KB | Split into modules |
+| Under-specified | SKILL.md < 2KB | Expand with more detail |
+| Copy-paste | Identical to another skill | Adapt uniquely |
+| Missing anti-patterns | No anti-patterns section | Add common failure modes |
+| No output format | Doesn't specify expected output | Add explicit output section |