@uluops/setup 0.4.0 → 0.6.0

package/assets/{commands/agents/code-validate.md → claude-code/commands/agents/validate.md}

@@ -1,135 +1,156 @@
----
-name: code-validate
-description: Validates code quality, standards compliance, and test coverage. Use after each implementation phase.
-model: sonnet
----
-
-# Code Validator
-Validates code quality, standards compliance, and test coverage. Use after each implementation phase.
-
-## Arguments
-
-**Usage:** `/agents:code-validate <directory>`
-
-**Examples:**
-- `/agents:code-validate ./services/auth-service`
-- `/agents:code-validate ./packages/api`
-- `/agents:code-validate .`
-
-**Target Directory:** $ARGUMENTS
-
----
-
-## Pre-Flight
-
-```bash
-echo "Running code validation on $ARGUMENTS..."
-echo "========================================"
-```
-
-Verify the target directory exists:
-
-```bash
-test -d "$ARGUMENTS" && echo "✓ Directory exists: $ARGUMENTS" || echo "ERROR: Directory '$ARGUMENTS' not found"
-```
-
-Enter and confirm location:
-
-```bash
-cd "$ARGUMENTS" && pwd
-```
-
-Check path exists:
-
-```bash
-[ -e "$ARGUMENTS" ] && echo "✓ $ARGUMENTS exists" || echo "Target directory does not exist"
-```
-
-
----
-
-## Agent Invocation
-
-Run the Code Validator agent on the validated target directory:
-
-**Agent:** code-validator-agent.md
-**Model:** Sonnet
-**Target:** $ARGUMENTS
-
-The agent performs code quality validation across 4 categories (100 points total):
-
-| Category | Points | Focus |
-|----------|--------|-------|
-| Code Quality | 30 | Function design, naming, duplication, error handling, complexity |
-| Standards Compliance | 25 | Style guide adherence, formatting, imports, documentation |
-| Testing | 25 | Unit tests, edge cases, behavior verification, test execution |
-| Best Practices | 20 | Security basics, performance, separation of concerns, dependencies |
-
----
-
-## Auto-Fail Conditions
-
-Critical issues that trigger immediate FAIL regardless of score:
-
-| ID | Condition |
-|----|-----------|
-| **AF-001** | Security vulnerabilities detected |
-| **AF-002** | Missing error handling in critical paths |
-| **AF-003** | Code does not function |
-| **AF-004** | Missing tests for core functionality |
-| **AF-005** | Breaking changes without migration path |
-
----
-
-## Decision Thresholds
-
-| Score | Decision | Meaning |
-|-------|----------|---------|
-| **>=70** | ✅ PASS | Validation passed, proceed to next phase |
-| **<70** | ❌ FAIL | Validation failed, fix issues before proceeding |
-
-**Note:** Any critical issue triggers FAIL regardless of score.
-
----
-
-
-## PERSIST TO TRACKER (Required)
-
-> **IMPORTANT:** Save to tracker IMMEDIATELY after agent completes, BEFORE presenting the summary to the user. The workflow is not complete until results are persisted.
-**1. Save to tracker (REQUIRED):**
-
-```bash
-agent-metrics buffer list --since 5m -f tracker
-```
-
-mcp__uluops-tracker__save_features_list
-
-**2. Verify saved:** Compare `json.summary.total_issues` with saved count.
-
-**3. THEN present summary to user.**
-
-
-### Field Mappings
-
-**From JSON OUTPUT to Tracker:**
-| Source | Tracker Field | Notes |
-|--------|---------------|-------|
-| `json.result.score` | `validators[].score` | Total score |
-| `json.result.decision` | `validators[].status` | PASS/FAIL |
-| `buffer.model` | `validators[].model` | From agent-metrics buffer |
-| `buffer.tokens.input_tokens` | `input_tokens` | Raw input tokens |
-| `buffer.tokens.output_tokens` | `output_tokens` | Output tokens |
-| `buffer.tokens.cache_creation_tokens` | `cache_creation_tokens` | Cache creation |
-| `buffer.tokens.cache_read_tokens` | `cache_read_tokens` | Cache reads |
-| `buffer.tokens.total_effective_tokens` | `total_effective_tokens` | Effective total |
-| `json.categories[].findings[].issues[]` | `recommendations[]` | Flatten nested structure |
-
-**Note:** `json` = agent's JSON OUTPUT, `buffer` = `agent-metrics buffer list -f tracker`
-
----
-
-## Source
-
-**CDL Schema:** `udl/definition-languages/cdl-schema-v1.1.0.json`
-**CDL Source:** `/home/alexs/uluops/uluops-agent-workflows/udl/cdl/v1/validate.command.yaml`
-**Agent:** `agents/code-validator-agent.md`
+---
+name: validate
+description: Validates code quality, standards compliance, and test coverage. Use after each implementation phase.
+---
+
+# Code Validator v1
+Validates code quality, standards compliance, and test coverage. Use after each implementation phase.
+
+## What's New in v1
+
+| Feature | Description |
+|---------|-------------|
+| **Calibration Examples** | Reference scenarios for consistent scoring |
+| **Failure Code Examples** | Worked examples mapping issues to taxonomy codes |
+| **Token Budget** | Output length guidance |
+| **Display IDs** | Auto-fail conditions have numbered IDs |
+
+## Arguments
+
+**Usage:** `/agents:validate <directory>`
+
+**Examples:**
+- `/agents:validate ./services/auth-service`
+- `/agents:validate ./packages/api`
+- `/agents:validate .`
+
+**Target Directory:** $ARGUMENTS
+
+
+---
+
+## Pre-Flight
+
+```bash
+echo "Running code validation on $ARGUMENTS..."
+echo "========================================"
+```
+
+Verify the target directory exists:
+
+```bash
+test -d "$ARGUMENTS" && echo "✓ Directory exists: $ARGUMENTS" || echo "ERROR: Directory '$ARGUMENTS' not found"
+```
+
+Enter and confirm location:
+
+```bash
+cd "$ARGUMENTS" && pwd
+```
+
+Check path exists:
+
+```bash
+[ -e "$ARGUMENTS" ] && echo "✓ $ARGUMENTS exists" || echo "Target directory does not exist"
+```
+
+
+---
+
+## Agent Invocation
+
+Run the Code Validator agent on the validated target directory:
+
+**Agent:** code-validator-agent.md
+**Model:** Sonnet
+**Target:** $ARGUMENTS
+
+
+---
+
+## Decision Thresholds
+
+| Score | Decision | Meaning |
+|-------|----------|---------|
+| **>=70** | ✅ PASS | Validation passed, proceed to next phase |
+| **<70** | ❌ FAIL | Validation failed, fix issues before proceeding |
+
+**Note:** Any critical issue triggers FAIL regardless of score.
+
+---
+
+## Post-Flight Actions
+
+### On Success
+
+Validation passed with score >= 70
+
+```bash
+exit 0
+```
+
+### On Failure
+
+Validation failed. Review issues above.
+
+```bash
+exit 1
+```
+
+
+---
+
+
+## PERSIST TO TRACKER (Required)
+
+> **IMPORTANT:** Save to tracker IMMEDIATELY after agent completes, BEFORE presenting the summary to the user. The workflow is not complete until results are persisted.
+**1. Save to tracker (REQUIRED):**
+
+```bash
+agent-metrics buffer list --since 5m -f tracker
+```
+
+mcp__uluops-tracker__save_run
+
+**2. Verify saved:** Compare `json.summary.total_issues` with saved count.
+
+**3. THEN present summary to user.**
+
+
+### Field Mappings
+
+**Definition identity (REQUIRED for execution tracking):**
+| Tracker Field | Value | Notes |
+|---------------|-------|-------|
+| `definition_type` | `command` | From CDL interface |
+| `definition_name` | `validate` | From CDL interface |
+| `definition_version` | `1.0.2` | From CDL interface |
+
+**From JSON OUTPUT to Tracker:**
+| Source | Tracker Field | Notes |
+|--------|---------------|-------|
+| `json.result.score` | `agents[].score` | Total score |
+| `json.result.decision` | `agents[].decision` | PASS/FAIL |
+| `buffer.model` | `validators[].model` | From agent-metrics buffer |
+| `buffer.tokens.input_tokens` | `input_tokens` | Raw input tokens |
+| `buffer.tokens.output_tokens` | `output_tokens` | Output tokens |
+| `buffer.tokens.cache_creation_tokens` | `cache_creation_tokens` | Cache creation |
+| `buffer.tokens.cache_read_tokens` | `cache_read_tokens` | Cache reads |
+| `buffer.tokens.total_effective_tokens` | `total_effective_tokens` | Effective total |
+| `json.categories[].findings[].issues[]` | `recommendations[]` | Flatten nested structure |
+| `json.analysis.records[]` | `analysis_records[]` | Structured analysis records (v1.4.0) |
+| `json.analysis.system_metrics` | `analysis_summary.system_metrics` | Agent-type-specific metrics |
+| `json.analysis.category_scores[]` | `analysis_summary.category_scores[]` | Category score breakdown |
+| `json.analysis.epistemic_assessment` | `analysis_summary.epistemic_assessment` | Failure signature risk ratings |
+| `json.analysis.audit_implications[]` | `analysis_summary.audit_implications[]` | Trajectory projections |
+
+**Note:** `json` = agent's JSON OUTPUT, `buffer` = `agent-metrics buffer list -f tracker`
+**Note:** `analysis_records` and `analysis_summary` are optional (v1.4.0). Omit if agent output has no `analysis` section.
+
+---
+
+## Source
+
+**CDL Schema:** `udl/definition-languages/cdl-schema-v1_3_0.json`
+**CDL Source:** `/Users/aself/uluops/uluops-agent-workflows/udl/cdl/v1/validate.command.yaml`
+**Agent:** `agents/code-validator-agent.md`

package/assets/claude-code/commands/agents/workflow-synthesis.md

@@ -0,0 +1,157 @@
+---
+name: workflow-synthesis
+description: Synthesizes cross-cutting insights from multiple upstream agent outputs. Identifies convergence, divergence, blind spots, and emergent patterns. Use at the end of any multi-agent workflow for integrated meta-analysis.
+---
+
+# Workflow Synthesis v1
+Synthesizes cross-cutting insights from multiple upstream agent outputs. Identifies convergence, divergence, blind spots, and emergent patterns. Use at the end of any multi-agent workflow for integrated meta-analysis.
+
+## What's New in v1
+
+| Feature | Description |
+|---------|-------------|
+| **Calibration Examples** | Reference scenarios for consistent scoring |
+| **Failure Code Examples** | Worked examples mapping issues to taxonomy codes |
+| **Token Budget** | Output length guidance |
+| **Display IDs** | Auto-fail conditions have numbered IDs |
+
+## Arguments
+
+**Usage:** `/agents:workflow-synthesis <directory>`
+
+**Examples:**
+- `/agents:workflow-synthesis uluops-registry-api/`
+- `/agents:workflow-synthesis ops-uluops-api/`
+- `/agents:workflow-synthesis udl/adl/v3/code-validator.agent.yaml`
+- `/agents:workflow-synthesis docs/specs/cognitive-lens-library-spec.md`
+
+**Target Directory:** $ARGUMENTS
+
+
+---
+
+## Pre-Flight
+
+```bash
+echo "Synthesizing cross-cutting insights from upstream agent outputs for $ARGUMENTS..."
+echo "================================================================================="
+```
+
+Verify the target directory exists:
+
+```bash
+test -d "$ARGUMENTS" && echo "✓ Directory exists: $ARGUMENTS" || echo "ERROR: Directory '$ARGUMENTS' not found"
+```
+
+Enter and confirm location:
+
+```bash
+cd "$ARGUMENTS" && pwd
+```
+
+Check path exists:
+
+```bash
+[ -e "$ARGUMENTS" ] && echo "✓ $ARGUMENTS exists" || echo "Target file or directory does not exist"
+```
+
+
+---
+
+## Agent Invocation
+
+Run the Workflow Synthesis agent on the validated target directory:
+
+**Agent:** workflow-synthesis-agent.md
+**Model:** Opus
+**Target:** $ARGUMENTS
+
+
+---
+
+## Decision Thresholds
+
+| Score | Decision | Meaning |
+|-------|----------|---------|
+| **>=75** | ✅ PASS | Validation passed, proceed to next phase |
+| **<75** | ❌ FAIL | Validation failed, fix issues before proceeding |
+
+**Note:** Any critical issue triggers FAIL regardless of score.
+
+---
+
+## Post-Flight Actions
+
+### On Success
+
+Synthesis INTEGRATED — cross-cutting insights produced from upstream agent outputs
+
+```bash
+exit 0
+```
+
+### On Failure
+
+Synthesis FRAGMENTED — upstream findings do not compose into emergent insights for this artifact
+
+```bash
+exit 1
+```
+
+
+---
+
+
+## PERSIST TO TRACKER (Required)
+
+> **IMPORTANT:** Save to tracker IMMEDIATELY after agent completes, BEFORE presenting the summary to the user. The workflow is not complete until results are persisted.
+**1. Get token metrics from buffer:**
+```bash
+agent-metrics buffer list --since 5m -f tracker
+```
+
+**2. Save to tracker (DO THIS FIRST):**
+
+mcp__uluops-tracker__save_run
+
+**3. Verify saved:** Compare `json.summary.total_issues` with saved count.
+
+**4. THEN present summary to user.**
+
+### Field Mappings
+
+**Definition identity (REQUIRED for execution tracking):**
+| Tracker Field | Value | Notes |
+|---------------|-------|-------|
+| `definition_type` | `command` | From CDL interface |
+| `definition_name` | `workflow-synthesis` | From CDL interface |
+| `definition_version` | `1.0.2` | From CDL interface |
+
+**From JSON OUTPUT to Tracker:**
+| Source | Tracker Field | Notes |
+|--------|---------------|-------|
+| `json.result.score` | `agents[].score` | Total score |
+| `json.result.decision` | `agents[].decision` | PASS/FAIL |
+| `buffer.model` | `validators[].model` | From agent-metrics buffer |
+| `buffer.tokens.input_tokens` | `input_tokens` | Raw input tokens |
+| `buffer.tokens.output_tokens` | `output_tokens` | Output tokens |
+| `buffer.tokens.cache_creation_tokens` | `cache_creation_tokens` | Cache creation |
+| `buffer.tokens.cache_read_tokens` | `cache_read_tokens` | Cache reads |
+| `buffer.tokens.total_effective_tokens` | `total_effective_tokens` | Effective total |
+| `json.categories[].findings[].issues[]` | `recommendations[]` | Flatten nested structure |
+| `json.analysis.records[]` | `analysis_records[]` | Structured analysis records (v1.4.0) |
+| `json.analysis.system_metrics` | `analysis_summary.system_metrics` | Agent-type-specific metrics |
+| `json.analysis.category_scores[]` | `analysis_summary.category_scores[]` | Category score breakdown |
+| `json.analysis.epistemic_assessment` | `analysis_summary.epistemic_assessment` | Failure signature risk ratings |
+| `json.analysis.audit_implications[]` | `analysis_summary.audit_implications[]` | Trajectory projections |
+
+**Note:** `json` = agent's JSON OUTPUT, `buffer` = `agent-metrics buffer list -f tracker`
+**Note:** `analysis_records` and `analysis_summary` are optional (v1.4.0). Omit if agent output has no `analysis` section.
+
+---
+
+## Source
+
+**CDL Schema:** `udl/definition-languages/cdl-schema-v1_3_0.json`
+**CDL Source:** `/Users/aself/uluops/uluops-agent-workflows/udl/cdl/v1/workflow-synthesis.command.yaml`
+**Agent:** `agents/workflow-synthesis-agent.md`

package/assets/{commands → claude-code/commands}/pipelines/aristotle.md

@@ -12,7 +12,7 @@ model: opus
 | Field | Value |
 |-------|-------|
 | Name | `aristotle-pipeline` |
-| Version | 1.0.0 |
+| Version | 1.0.1 |
 | Domain | software |
 | Subdomain | cognitive-lens |
 | Tags | `cognitive-lens`, `aristotle`, `philosophy`, `sequential` |
@@ -98,18 +98,18 @@ explore (Categorical Mapping)
 
 ### Tracker Persistence
 
-After all stages complete, save results to the tracker using `save_features_list`:
+After all stages complete, save results to the tracker using `save_run`:
 
 - **definition_type:** `pipeline`
 - **definition_name:** `aristotle-pipeline`
-- **definition_version:** `1.0.0`
+- **definition_version:** `1.0.1`
 - **workflow_type:** `cognitive-lens`
 - **project:** `$ARGUMENTS` (the target project name)
-- **validators:** Collect each validator/agent result with name, score, status, model, and token usage
-- **recommendations:** Collect ALL issues from ALL stages into a single recommendations array with validator, title, priority, severity, failure_code, file_path, line_number, description, and type
-- **summary:** `{ all_gates_passed: <true if all abort-gates passed>, average_score: <mean of all validator scores> }`
+- **agents:** Collect each agent result with name, score, decision, model, and tokens
+- **recommendations:** Collect ALL issues from ALL stages into a single recommendations array with agent, title, priority, severity, failure_code, file_path, line_number, description, and type
+- **summary:** `{ all_gates_passed: <true if all abort-gates passed>, average_score: <mean of all agent scores> }`
 
-This MUST be a single bulk call — do NOT create individual issues. The `save_features_list` tool auto-increments the run number and detects regressions from prior runs.
+This MUST be a single bulk call — do NOT create individual issues. The `save_run` tool auto-increments the run number and detects regressions from prior runs.
 
 **Token Metrics:**
 
@@ -140,4 +140,4 @@ Map buffer fields to tracker: `input_tokens`, `output_tokens`, `cache_creation_t
 - **POTENTIAL — Significant ontological gaps remain.**
 
 ---
-*Generated from PDL v1.0.0 | Pipeline: aristotle-pipeline*
+*Generated from PDL v1.0.1 | Pipeline: aristotle-pipeline*

package/assets/{commands → claude-code/commands}/pipelines/ship.md

@@ -12,7 +12,7 @@ model: opus
 | Field | Value |
 |-------|-------|
 | Name | `ship` |
-| Version | 1.1.0 |
+| Version | 1.1.1 |
 | Domain | software |
 | Subdomain | release |
 | Tags | `ship`, `validation`, `release`, `ci-cd` |
@@ -143,18 +143,18 @@ preflight (Pre-Flight Detection) [parallel]
 
 ### Tracker Persistence
 
-After all stages complete, save results to the tracker using `save_features_list`:
+After all stages complete, save results to the tracker using `save_run`:
 
 - **definition_type:** `pipeline`
 - **definition_name:** `ship`
-- **definition_version:** `1.1.0`
+- **definition_version:** `1.1.1`
 - **workflow_type:** `ship`
 - **project:** `$ARGUMENTS` (the target project name)
-- **validators:** Collect each validator/agent result with name, score, status, model, and token usage
-- **recommendations:** Collect ALL issues from ALL stages into a single recommendations array with validator, title, priority, severity, failure_code, file_path, line_number, description, and type
-- **summary:** `{ all_gates_passed: <true if all abort-gates passed>, average_score: <mean of all validator scores> }`
+- **agents:** Collect each agent result with name, score, decision, model, and tokens
+- **recommendations:** Collect ALL issues from ALL stages into a single recommendations array with agent, title, priority, severity, failure_code, file_path, line_number, description, and type
+- **summary:** `{ all_gates_passed: <true if all abort-gates passed>, average_score: <mean of all agent scores> }`
 
-This MUST be a single bulk call — do NOT create individual issues. The `save_features_list` tool auto-increments the run number and detects regressions from prior runs.
+This MUST be a single bulk call — do NOT create individual issues. The `save_run` tool auto-increments the run number and detects regressions from prior runs.
 
 **Token Metrics:**
 
@@ -185,4 +185,4 @@ Map buffer fields to tracker: `input_tokens`, `output_tokens`, `cache_creation_t
 - **NOT READY — Blocking issues found. Fix and re-run.**
 
 ---
-*Generated from PDL v1.1.0 | Pipeline: ship*
+*Generated from PDL v1.1.1 | Pipeline: ship*

package/assets/claude-code/commands/workflows/post-implementation.md

@@ -0,0 +1,60 @@
+---
+name: post-implementation
+description: Iterative validation workflow. Run after each implementation phase until all agents pass. Includes code optimization and optional TypeScript, MCP, and frontend validators. Use --frontend flag for React/Tailwind projects.
+tools: Read, Grep, Glob, Bash
+model: opus
+---
+
+# Post-Implementation
+
+> Iterative validation workflow. Run after each implementation phase until all agents pass. Includes code optimization and optional TypeScript, MCP, and frontend validators. Use --frontend flag for React/Tailwind projects.
+
+Duration: 6-15 minutes
+**Arguments**: `target`
+## Pre-Flight Detection
+
+- **typescript_detected**: `test -f '{{ target }}/tsconfig.json'`
+- **mcp_python_detected**: `grep -rqE --include="*.py" --exclude-dir="node_modules" --exclude-dir="dist" --exclude-dir="__pycache__" --exclude-dir=".venv" "from mcp|import mcp|FastMCP|@mcp\." {{ target }} 2>/dev/null`
+- **mcp_typescript_detected**: `grep -rqE --include="*.ts" --include="*.js" --exclude-dir="node_modules" --exclude-dir="dist" "from .@modelcontextprotocol/sdk.|new McpServer|StdioServerTransport" {{ target }} 2>/dev/null`
+- **mcp_dependency_detected**: `grep -rqE --include="package.json" --include="pyproject.toml" --include="requirements.txt" --exclude-dir="node_modules" --exclude-dir="dist" "@modelcontextprotocol|\"mcp\":" {{ target }} 2>/dev/null`
+- **frontend_detected**: `test -n "$(find {{ target }} \( -name "*.tsx" -o -name "*.jsx" \) ! -path "*/node_modules/*" ! -path "*/dist/*" -print -quit 2>/dev/null)"`
+## Execution
+
+Ask user: **Sequential** (stop on first failure) or **Parallel** (run groups concurrently)?
+```
+Group 1 (gate): validate
+Group 2 (parallel): type-safety + mcp-validator + test-architect
+Group 3 (sequential): optimizer
+Group 4 (parallel): public-interface + frontend
+Group 5 (sequential): security
+```
+## Phases
+
+| # | Agent | Threshold | Gate | Condition |
+|---|-------|-----------|------|-----------|
+| 1 | validate@latest | threshold >= 70, on fail: stop | stop | — |
+| 2 | type-safety@latest | threshold >= 75, on fail: stop | stop | context.typescript_detected |
+| 3 | mcp-validate@latest | threshold >= 80, warn if < 65, on fail: stop | stop | context.mcp_detected |
+| 4 | test-review@latest | threshold >= 70, on fail: warn | stop | — |
+| 5 | optimize@latest | threshold >= 75, on fail: warn | stop | — |
+| 6 | public-interface@latest | threshold >= 75, on fail: warn | stop | — |
+| 7 | frontend@latest | threshold >= 75, on fail: warn | stop | context.frontend_enabled |
+| 8 | security@latest | threshold >= 75, on fail: warn | stop | — |
+
+**Code Validation**: Correctness and best practices; Error handling patterns
+**TypeScript Validation** (after validate): Explicit any usage and type holes; Public API type quality
+**MCP Protocol Validation** (after validate): Tools, resources, and prompts implementation; Transport and security compliance
+**Test Architecture Review** (after validate): Test quality, not just coverage; Critical path coverage
+**Code Optimization** (after test-architect): Duplication and structure; Hot path performance
+**Public Interface Validation** (after optimizer): README accuracy and completeness; Export hygiene
+**Frontend Validation** (after optimizer): Accessibility and theme consistency; Component quality and performance
+**Security Analysis** (after public-interface): OWASP Top 10 and CWE Top 25 coverage; Input validation and injection prevention
+## Scoring
+
+**Method**: weighted_average
+ — validate: 20%, type-safety: 15%, mcp-validator: 10%, test-architect: 15%, optimizer: 10%, public-interface: 10%, frontend: 10%, security: 10%
+## Results Submission
+
+Write markdown report to: `{{ target_path }}/{{ features_file }}`
+Save ALL findings to tracker via `mcp_uluops-tracker_save_run` with project=`{{ target_name }}`, workflow_type=`post-implementation`, definition_type=`workflow`, definition_name=`post-implementation`, definition_version=`3.2.1`. Include agents array (name, score, decision, model, tokens) and recommendations array (agent, title, priority, severity, description, file_path, line_number). Each file:line reference becomes a separate recommendation. Priority: blocking=critical, warnings=suggested, post-ship=backlog.
+After saving, query tracker and compare counts. Mismatches from cross-phase deduplication are expected — warn only, do not re-attempt.

package/assets/claude-code/commands/workflows/pre-implementation.md

@@ -0,0 +1,46 @@
+---
+name: pre-implementation
+description: Validates proposed design and architecture BEFORE implementation begins. Runs architect review plus optional docs validation. Blocks implementation if design has critical flaws.
+tools: Read, Grep, Glob, Bash
+model: opus
+---
+
+# Pre-Implementation
+
+> Validates proposed design and architecture BEFORE implementation begins. Runs architect review plus optional docs validation. Blocks implementation if design has critical flaws.
+
+Duration: 3-10 minutes
+**Arguments**: `target` (required)
+## Pre-Flight Detection
+
+- **design_docs_detected**: `test -n "$(find . \( -name '*design*' -o -name '*spec*' -o -name '*prd*' -o -name '*plan*' -o -name 'README.md' \) ! -path '*/node_modules/*' ! -path '*/dist/*' ! -path '*/.venv/*' -print -quit 2>/dev/null)"`
+## Execution
+
+Ask user: **Sequential** (stop on first failure) or **Parallel** (run groups concurrently)?
+```
+Group 1 (parallel): architect + docs-validator
+Group 2 (sequential): assumption-review
+Group 3 (always): persist
+```
+## Phases
+
+| # | Agent | Threshold | Gate | Condition |
+|---|-------|-----------|------|-----------|
+| 1 |  | threshold >= 75, on fail: stop | stop | — |
+| 2 |  | threshold >= 75, on fail: warn | stop | context.has_design_docs |
+| 3 |  | threshold >= 70, on fail: warn | stop | — |
+| 4 | workflow-synthesis@latest | threshold >= 0, on fail: warn | stop | — |
+
+**Architecture Review**: Architectural fit with existing codebase; Design quality and complexity; Scope bounds (LOC, files, dependencies); Completeness of error handling and data flow
+**Documentation Validation**: Design doc completeness; API contract clarity; Data flow documentation
+**Epistemic Assumption Review** (after architect): Implicit assumptions in proposed design; Overclaiming of capabilities or guarantees; Reasoning quality behind architectural decisions
+**Persist Results**: Synthesize findings and persist to tracker
+## Scoring
+
+**Method**: weighted_average
+ — architect: 55%, assumption-review: 20%, docs-validator: 25%
+## Results Submission
+
+Write markdown report to: `{{ target_path }}/{{ report_file }}`
+Save ALL findings to tracker via `mcp_uluops-tracker_save_run` with project=`{{ target_name }}`, workflow_type=`pre-implementation`, definition_type=`workflow`, definition_name=`pre-implementation`, definition_version=`2.0.1`. Include agents array (name, score, decision, model, tokens) and recommendations array (agent, title, priority, severity, description, file_path, line_number). Each file:line reference becomes a separate recommendation. Priority: blocking=critical, warnings=suggested, post-ship=backlog.
+After saving, query tracker and compare counts. Mismatches from cross-phase deduplication are expected — warn only, do not re-attempt.

package/assets/{commands → claude-code/commands}/workflows/prompt-audit.md

@@ -14,7 +14,7 @@ Duration: 6-15 minutes
 ## Pre-Flight Detection
 
 - **is_agent_definition**: `grep -rqE --include='{{ file }}'  '^(name|description|type):' . 2>/dev/null`
-- **is_agent_command**: `grep -rqE --include='{{ file }}'  'subagent_type|\[validator:' . 2>/dev/null`
+- **is_agent_command**: `grep -rqE --include='{{ file }}'  'subagent_type|\[agent:' . 2>/dev/null`
 ## Execution
 
 ```
@@ -40,5 +40,5 @@ Group 3 (sequential): prompt-quality
 ## Results Submission
 
 Write markdown report to: `{{ target_path }}/{{ report_file }}`
-Save ALL findings to tracker via `mcp__uluops-tracker__save_features_list` with project=`{{ target_name }}`, workflow_type=`prompt-audit`, definition_type=`workflow`, definition_name=`prompt-audit`, definition_version=`2.0.0`. Include validators array (name, score, status, model) and recommendations array (validator, title, priority, severity, description, file_path, line_number). Each file:line reference becomes a separate recommendation. Priority: blocking=critical, warnings=suggested, post-ship=backlog.
+Save ALL findings to tracker via `mcp_uluops-tracker_save_run` with project=`{{ target_name }}`, workflow_type=`prompt-audit`, definition_type=`workflow`, definition_name=`prompt-audit`, definition_version=`2.0.1`. Include agents array (name, score, decision, model, tokens) and recommendations array (agent, title, priority, severity, description, file_path, line_number). Each file:line reference becomes a separate recommendation. Priority: blocking=critical, warnings=suggested, post-ship=backlog.
 After saving, query tracker and compare counts. Mismatches from cross-phase deduplication are expected — warn only, do not re-attempt.