ai-eng-system 0.2.1 → 0.2.3

package/README.md

@@ -65,6 +65,8 @@ See [docs/spec-driven-workflow.md](./docs/spec-driven-workflow.md) for complete
 
 > **🔄 Ralph Wiggum**: Add `--ralph` flag to any phase command for persistent iteration. Perfect for complex tasks requiring multiple refinement cycles. [Learn more →](./docs/ralph-wiggum-integration-complete.md)
 
+> **🛠️ Fixed (v0.2.2)**: Ralph Wiggum now runs continuously without stopping between phases. Full backward compatibility maintained. [Fix details →](./docs/ralph-wiggum-autonomous-looping-fix.md)
+
 ### Additional Commands (13 total)
 - `/ai-eng/optimize` - Prompt enhancement (+45% quality)
 - `/ai-eng/deploy` - Pre-deployment checklists

package/dist/.claude-plugin/commands/plan.md

@@ -493,25 +493,11 @@ Successful planning achieves:
 - ✅ Dependencies are clearly documented
 - ✅ All spec acceptance criteria are covered (if spec exists)
 - ✅ Supporting artifacts generated (data-model, contracts)
-- ✅ Risk assessment completed
-- ✅ Testing strategy defined
-- ✅ Ready to feed into `/ai-eng/work`
-
-## Execution
-
-After planning, execute the plan using:
-
-```bash
-bun run scripts/run-command.ts plan "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts plan "implement auth" --from-spec=specs/auth/spec.md --output=plans/auth.yaml`
-- `bun run scripts/run-command.ts plan --from-research=docs/research/auth.md --scope=implementation`
-
-After creating the plan, rate your confidence in its completeness and accuracy (0.0-1.0). Identify any uncertainties about task decomposition, missing dependencies, or areas where acceptance criteria may be ambiguous. Note any implementation risks that weren't adequately addressed in the plan.
-
-## Ralph Wiggum Iteration Mode
+ - ✅ Risk assessment completed
+ - ✅ Testing strategy defined
+ - ✅ Ready to feed into `/ai-eng/work`
+ 
+ ## Ralph Wiggum Iteration Mode
 
 When `--ralph` flag is enabled, the planning process follows a persistent refinement cycle:
 

package/dist/.claude-plugin/commands/ralph-wiggum.md

@@ -65,6 +65,21 @@ I bet you can't manage continuous iteration across all workflow phases without g
 /ai-eng/ralph-wiggum "user authentication" --dry-run
 ```
 
+## How This Command Works
+
+This command is a **self-orchestrating prompt** - when you invoke `/ai-eng/ralph-wiggum`, the AI reads these instructions and executes each phase directly within the conversation.
+
+**Execution Model:**
+1. **NOT a CLI script** - There is no external runner. The AI IS the executor.
+2. **Phase commands are internal** - When instructions say "Execute `/ai-eng/research`", the AI follows that command's instructions directly (not spawning a subprocess).
+3. **Continuous flow** - The AI proceeds through all phases autonomously unless checkpoints are set.
+
+**To run this command:**
+- Claude Code: Type `/ai-eng/ralph-wiggum "your feature description"`
+- OpenCode: Type `/ai-eng/ralph-wiggum "your feature description"`
+
+The AI will then follow all phases in this document from start to finish.
+
 ## Options
 
 | Option | Type | Default | Description |
@@ -85,6 +100,7 @@ I bet you can't manage continuous iteration across all workflow phases without g
 | `--dry-run` | boolean | false | Show plan without executing |
 | `--resume` | boolean | false | Resume from last checkpoint |
 | `--parallel` | boolean | false | Run phases in parallel where safe |
+| `--refine-each-phase` | boolean | false | Re-invoke prompt-refinement at each phase (interactive mode) |
 
 **Default Completion Promise:**
 ```
@@ -95,6 +111,18 @@ I bet you can't manage continuous iteration across all workflow phases without g
 
 Load `skills/prompt-refinement/SKILL.md` and use phase: `plan` to transform your prompt into structured TCRO format (Task, Context, Requirements, Output). Ask clarifying questions if feature description, acceptance criteria, technical approach, or quality gates are unclear.
 
+**Store the refined output as `$REFINED_CONTEXT`** - this will be used for ALL subsequent phases without re-invoking prompt-refinement.
+
+### Autonomous Execution Model
+
+After Phase 0 completes:
+1. **Context is locked**: The refined TCRO applies to ALL subsequent phases
+2. **No re-refinement**: Do NOT invoke prompt-refinement again (unless `--refine-each-phase` is set)
+3. **Continuous execution**: Proceed through phases without pausing for user input
+4. **Checkpoints are optional**: Only pause if `--checkpoint` flag is explicitly set
+
+This enables overnight/unattended execution as intended by the Ralph Wiggum pattern.
+
 ## Phase 1: Git Setup
 
 ### 1.1 Create Feature Branch
@@ -137,7 +165,7 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Goal**: Gather context, patterns, and technical guidance
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -157,11 +185,13 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Checkpoint** (if `--checkpoint=all`): Pause and ask for approval before proceeding
 
+**→ CONTINUE**: Immediately proceed to Phase 2.3 (Specify) without waiting for user input.
+
 ### 2.3 Specify Phase (if not skipped)
 
 **Goal**: Create detailed specification with user stories and acceptance criteria
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -186,11 +216,13 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Checkpoint** (if `--checkpoint=all`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 2.4 (Plan) without waiting for user input.
+
 ### 2.4 Plan Phase (if not skipped)
 
 **Goal**: Create implementation plan with tasks and dependencies
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -215,11 +247,13 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Checkpoint** (if `--checkpoint=all`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 2.5 (Work) without waiting for user input.
+
 ### 2.5 Work Phase
 
 **Goal**: Implement feature with TDD, quality gates, and iterative refinement
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -248,11 +282,13 @@ npm test && npm run lint && npm run build
 
 **Checkpoint** (if `--checkpoint=all` or `--checkpoint=review`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 2.6 (Review) without waiting for user input.
+
 ### 2.6 Review Phase
 
 **Goal**: Comprehensive multi-perspective code review
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -271,6 +307,8 @@ npm test && npm run lint && npm run build
 
 **Checkpoint** (if `--checkpoint=all` or `--checkpoint=review`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 3 (Gap Analysis) without waiting for user input.
+
 ## Phase 3: Gap Analysis
 
 ### 3.1 Analyze Review Findings
@@ -354,6 +392,12 @@ Based on gap analysis from 3.1:
   "max_cycles": 5,
   "current_phase": "WORK",
   "completed_phases": ["RESEARCH", "SPECIFY", "PLAN"],
+  "refined_context": {
+    "task": "Specific, actionable task statement",
+    "context": "Broader system, goals, constraints from CLAUDE.md",
+    "requirements": ["Must-have requirement 1", "Must-have requirement 2"],
+    "output": "What should be delivered (working code, tests, docs, PR)"
+  },
   "artifacts": {
     "research": "docs/research/2026-01-05-user-authentication.md",
     "spec": "specs/user-authentication/spec.md",
@@ -739,22 +783,8 @@ Successful build achieves:
 - ✅ Checkpoints saved and can be loaded
 - ✅ Progress displayed at appropriate verbosity
 - ✅ Safety limits enforced (max cycles, stuck detection)
-- ✅ Token usage tracked and displayed
-- ✅ Draft PR created with comprehensive summary
-- ✅ Build report generated with all artifacts and metrics
-
----
-
-## Execution
-
-After creating this command, the agent can invoke it with:
-
-```bash
-bun run scripts/run-command.ts ralph-wiggum "feature description" [options]
-```
-
-Examples:
-- `bun run scripts/run-command.ts ralph-wiggum "implement user auth" --checkpoint=all --verbose`
-- `bun run scripts/run-command.ts ralph-wiggum "API caching" --from-spec=specs/cache/spec.md --resume`
+ - ✅ Token usage tracked and displayed
+ - ✅ Draft PR created with comprehensive summary
+ - ✅ Build report generated with all artifacts and metrics
 
-After completing the full cycle, rate your confidence in feature completeness and quality (0.0-1.0). Identify any uncertainties about phase transitions, iterations that were inefficient, or areas where quality gates may have been too lenient or strict. Note any workflow improvements that could enhance future cycles.
+ ---

package/dist/.claude-plugin/commands/review.md

@@ -115,24 +115,12 @@ For each finding provide:
 
 ## Summary
 
+
 End with overall assessment: APPROVE, CHANGES_REQUESTED, or NEEDS_DISCUSSION.
 
 After completing review, rate your confidence in findings comprehensiveness (0.0-1.0). Identify any uncertainties about severity classifications, areas where review coverage may have been insufficient, or assumptions about code context. Note any perspectives that should have been applied or findings that may be false positives.
-
-## Execution
-
-Run a review using:
-
-```bash
-bun run scripts/run-command.ts review "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts review "src/" --type=security --severity=high --output=security-review.json`
-- `bun run scripts/run-command.ts review "." --focus=performance --verbose`
-
-## Ralph Wiggum Iteration Mode
-
+ 
+ ## Ralph Wiggum Iteration Mode
 When `--ralph` flag is enabled, the review process follows a persistent refinement cycle:
 
 ### Ralph Wiggum Review Cycle

package/dist/.claude-plugin/commands/specify.md

@@ -288,28 +288,14 @@ Always consider these categories:
 ## Success Criteria
 
 Successful specification achieves:
-- ✅ Well-structured user stories with acceptance criteria
-- ✅ All non-functional requirements defined
-- ✅ Ambiguities marked with [NEEDS CLARIFICATION]
-- ✅ Aligned with project philosophy (CLAUDE.md)
-- ✅ Ready to feed into `/ai-eng/plan`
-- ✅ User reviewed and approved
-
-## Execution
-
-After specification, create a plan using:
-
-```bash
-bun run scripts/run-command.ts specify "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts specify "user auth" --from-research=docs/research/auth.md --output=specs/auth/spec.md`
-- `bun run scripts/run-command.ts specify "payment system" --template=api --verbose`
-
-After creating specification, rate your confidence in its completeness and clarity (0.0-1.0). Identify any uncertainties about user requirements, areas where acceptance criteria may be ambiguous, or constraints that weren't adequately considered. Note any open questions or clarifications needed from stakeholders.
-
-## Integration
+ - ✅ Well-structured user stories with acceptance criteria
+ - ✅ All non-functional requirements defined
+ - ✅ Ambiguities marked with [NEEDS CLARIFICATION]
+ - ✅ Aligned with project philosophy (CLAUDE.md)
+ - ✅ Ready to feed into `/ai-eng/plan`
+ - ✅ User reviewed and approved
+ 
+ ## Integration
 
 - Can use output from `/ai-eng/research` via `--from-research`
 - Feeds into `/ai-eng/plan` for implementation planning

package/dist/.claude-plugin/commands/work.md

@@ -573,29 +573,14 @@ This will:
 ## Success Criteria
 
 A work session is successful when:
-- ✅ All tasks completed
-- ✅ All quality gates passed
-- ✅ All tests passing
-- ✅ Build succeeds
-- ✅ PR created and reviewed
-- ✅ Code merged to main
-
-## Execution
-
-Execute a plan using:
-
-```bash
-bun run scripts/run-command.ts work "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts work "specs/auth/plan.yaml" --verbose`
-- `bun run scripts/run-command.ts work --from-plan=plans/auth.yaml --continue`
-- `bun run scripts/run-command.ts work "FEAT-001" --dry-run`
-
-After completing implementation, rate your confidence in code quality and completeness (0.0-1.0). Identify any uncertainties about test coverage, areas where quality gates may have been too lenient, or implementation details that may need follow-up. Note any technical debt or known issues that should be tracked.
-
-## Integration
+ - ✅ All tasks completed
+ - ✅ All quality gates passed
+ - ✅ All tests passing
+ - ✅ Build succeeds
+ - ✅ PR created and reviewed
+ - ✅ Code merged to main
+ 
+ ## Integration
 
 - Reads from `/ai-eng/plan` output (plan.yaml)
 - Validates against `/ai-eng/specify` output (spec.md)

package/dist/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-eng-system",
-  "version": "0.2.1",
+  "version": "0.2.3",
   "description": "AI Engineering System with context engineering and research orchestration for Claude Code",
   "author": {
     "name": "v1truv1us"

package/dist/.opencode/command/ai-eng/plan.md

@@ -493,25 +493,11 @@ Successful planning achieves:
 - ✅ Dependencies are clearly documented
 - ✅ All spec acceptance criteria are covered (if spec exists)
 - ✅ Supporting artifacts generated (data-model, contracts)
-- ✅ Risk assessment completed
-- ✅ Testing strategy defined
-- ✅ Ready to feed into `/ai-eng/work`
-
-## Execution
-
-After planning, execute the plan using:
-
-```bash
-bun run scripts/run-command.ts plan "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts plan "implement auth" --from-spec=specs/auth/spec.md --output=plans/auth.yaml`
-- `bun run scripts/run-command.ts plan --from-research=docs/research/auth.md --scope=implementation`
-
-After creating the plan, rate your confidence in its completeness and accuracy (0.0-1.0). Identify any uncertainties about task decomposition, missing dependencies, or areas where acceptance criteria may be ambiguous. Note any implementation risks that weren't adequately addressed in the plan.
-
-## Ralph Wiggum Iteration Mode
+ - ✅ Risk assessment completed
+ - ✅ Testing strategy defined
+ - ✅ Ready to feed into `/ai-eng/work`
+ 
+ ## Ralph Wiggum Iteration Mode
 
 When `--ralph` flag is enabled, the planning process follows a persistent refinement cycle:
 

package/dist/.opencode/command/ai-eng/ralph-wiggum.md

@@ -65,6 +65,21 @@ I bet you can't manage continuous iteration across all workflow phases without g
 /ai-eng/ralph-wiggum "user authentication" --dry-run
 ```
 
+## How This Command Works
+
+This command is a **self-orchestrating prompt** - when you invoke `/ai-eng/ralph-wiggum`, the AI reads these instructions and executes each phase directly within the conversation.
+
+**Execution Model:**
+1. **NOT a CLI script** - There is no external runner. The AI IS the executor.
+2. **Phase commands are internal** - When instructions say "Execute `/ai-eng/research`", the AI follows that command's instructions directly (not spawning a subprocess).
+3. **Continuous flow** - The AI proceeds through all phases autonomously unless checkpoints are set.
+
+**To run this command:**
+- Claude Code: Type `/ai-eng/ralph-wiggum "your feature description"`
+- OpenCode: Type `/ai-eng/ralph-wiggum "your feature description"`
+
+The AI will then follow all phases in this document from start to finish.
+
 ## Options
 
 | Option | Type | Default | Description |
@@ -85,6 +100,7 @@ I bet you can't manage continuous iteration across all workflow phases without g
 | `--dry-run` | boolean | false | Show plan without executing |
 | `--resume` | boolean | false | Resume from last checkpoint |
 | `--parallel` | boolean | false | Run phases in parallel where safe |
+| `--refine-each-phase` | boolean | false | Re-invoke prompt-refinement at each phase (interactive mode) |
 
 **Default Completion Promise:**
 ```
@@ -95,6 +111,18 @@ I bet you can't manage continuous iteration across all workflow phases without g
 
 Load `skills/prompt-refinement/SKILL.md` and use phase: `plan` to transform your prompt into structured TCRO format (Task, Context, Requirements, Output). Ask clarifying questions if feature description, acceptance criteria, technical approach, or quality gates are unclear.
 
+**Store the refined output as `$REFINED_CONTEXT`** - this will be used for ALL subsequent phases without re-invoking prompt-refinement.
+
+### Autonomous Execution Model
+
+After Phase 0 completes:
+1. **Context is locked**: The refined TCRO applies to ALL subsequent phases
+2. **No re-refinement**: Do NOT invoke prompt-refinement again (unless `--refine-each-phase` is set)
+3. **Continuous execution**: Proceed through phases without pausing for user input
+4. **Checkpoints are optional**: Only pause if `--checkpoint` flag is explicitly set
+
+This enables overnight/unattended execution as intended by the Ralph Wiggum pattern.
+
 ## Phase 1: Git Setup
 
 ### 1.1 Create Feature Branch
@@ -137,7 +165,7 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Goal**: Gather context, patterns, and technical guidance
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -157,11 +185,13 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Checkpoint** (if `--checkpoint=all`): Pause and ask for approval before proceeding
 
+**→ CONTINUE**: Immediately proceed to Phase 2.3 (Specify) without waiting for user input.
+
 ### 2.3 Specify Phase (if not skipped)
 
 **Goal**: Create detailed specification with user stories and acceptance criteria
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -186,11 +216,13 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Checkpoint** (if `--checkpoint=all`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 2.4 (Plan) without waiting for user input.
+
 ### 2.4 Plan Phase (if not skipped)
 
 **Goal**: Create implementation plan with tasks and dependencies
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -215,11 +247,13 @@ This is the core Ralph Wiggum pattern - continuous iteration until completion.
 
 **Checkpoint** (if `--checkpoint=all`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 2.5 (Work) without waiting for user input.
+
 ### 2.5 Work Phase
 
 **Goal**: Implement feature with TDD, quality gates, and iterative refinement
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -248,11 +282,13 @@ npm test && npm run lint && npm run build
 
 **Checkpoint** (if `--checkpoint=all` or `--checkpoint=review`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 2.6 (Review) without waiting for user input.
+
 ### 2.6 Review Phase
 
 **Goal**: Comprehensive multi-perspective code review
 
-**Load Ralph Wiggum skill**: `skills/workflow/ralph-wiggum/SKILL.md`
+**Context**: Use `$REFINED_CONTEXT` from Phase 0. Do NOT re-invoke prompt-refinement unless `--refine-each-phase` is set.
 
 **Execute**:
 ```bash
@@ -271,6 +307,8 @@ npm test && npm run lint && npm run build
 
 **Checkpoint** (if `--checkpoint=all` or `--checkpoint=review`): Pause and ask for approval
 
+**→ CONTINUE**: Immediately proceed to Phase 3 (Gap Analysis) without waiting for user input.
+
 ## Phase 3: Gap Analysis
 
 ### 3.1 Analyze Review Findings
@@ -354,6 +392,12 @@ Based on gap analysis from 3.1:
   "max_cycles": 5,
   "current_phase": "WORK",
   "completed_phases": ["RESEARCH", "SPECIFY", "PLAN"],
+  "refined_context": {
+    "task": "Specific, actionable task statement",
+    "context": "Broader system, goals, constraints from CLAUDE.md",
+    "requirements": ["Must-have requirement 1", "Must-have requirement 2"],
+    "output": "What should be delivered (working code, tests, docs, PR)"
+  },
   "artifacts": {
     "research": "docs/research/2026-01-05-user-authentication.md",
     "spec": "specs/user-authentication/spec.md",
@@ -739,22 +783,8 @@ Successful build achieves:
 - ✅ Checkpoints saved and can be loaded
 - ✅ Progress displayed at appropriate verbosity
 - ✅ Safety limits enforced (max cycles, stuck detection)
-- ✅ Token usage tracked and displayed
-- ✅ Draft PR created with comprehensive summary
-- ✅ Build report generated with all artifacts and metrics
-
----
-
-## Execution
-
-After creating this command, the agent can invoke it with:
-
-```bash
-bun run scripts/run-command.ts ralph-wiggum "feature description" [options]
-```
-
-Examples:
-- `bun run scripts/run-command.ts ralph-wiggum "implement user auth" --checkpoint=all --verbose`
-- `bun run scripts/run-command.ts ralph-wiggum "API caching" --from-spec=specs/cache/spec.md --resume`
+ - ✅ Token usage tracked and displayed
+ - ✅ Draft PR created with comprehensive summary
+ - ✅ Build report generated with all artifacts and metrics
 
-After completing the full cycle, rate your confidence in feature completeness and quality (0.0-1.0). Identify any uncertainties about phase transitions, iterations that were inefficient, or areas where quality gates may have been too lenient or strict. Note any workflow improvements that could enhance future cycles.
+ ---

package/dist/.opencode/command/ai-eng/review.md

@@ -115,24 +115,12 @@ For each finding provide:
 
 ## Summary
 
+
 End with overall assessment: APPROVE, CHANGES_REQUESTED, or NEEDS_DISCUSSION.
 
 After completing review, rate your confidence in findings comprehensiveness (0.0-1.0). Identify any uncertainties about severity classifications, areas where review coverage may have been insufficient, or assumptions about code context. Note any perspectives that should have been applied or findings that may be false positives.
-
-## Execution
-
-Run a review using:
-
-```bash
-bun run scripts/run-command.ts review "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts review "src/" --type=security --severity=high --output=security-review.json`
-- `bun run scripts/run-command.ts review "." --focus=performance --verbose`
-
-## Ralph Wiggum Iteration Mode
-
+ 
+ ## Ralph Wiggum Iteration Mode
 When `--ralph` flag is enabled, the review process follows a persistent refinement cycle:
 
 ### Ralph Wiggum Review Cycle

package/dist/.opencode/command/ai-eng/specify.md

@@ -288,28 +288,14 @@ Always consider these categories:
 ## Success Criteria
 
 Successful specification achieves:
-- ✅ Well-structured user stories with acceptance criteria
-- ✅ All non-functional requirements defined
-- ✅ Ambiguities marked with [NEEDS CLARIFICATION]
-- ✅ Aligned with project philosophy (CLAUDE.md)
-- ✅ Ready to feed into `/ai-eng/plan`
-- ✅ User reviewed and approved
-
-## Execution
-
-After specification, create a plan using:
-
-```bash
-bun run scripts/run-command.ts specify "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts specify "user auth" --from-research=docs/research/auth.md --output=specs/auth/spec.md`
-- `bun run scripts/run-command.ts specify "payment system" --template=api --verbose`
-
-After creating specification, rate your confidence in its completeness and clarity (0.0-1.0). Identify any uncertainties about user requirements, areas where acceptance criteria may be ambiguous, or constraints that weren't adequately considered. Note any open questions or clarifications needed from stakeholders.
-
-## Integration
+ - ✅ Well-structured user stories with acceptance criteria
+ - ✅ All non-functional requirements defined
+ - ✅ Ambiguities marked with [NEEDS CLARIFICATION]
+ - ✅ Aligned with project philosophy (CLAUDE.md)
+ - ✅ Ready to feed into `/ai-eng/plan`
+ - ✅ User reviewed and approved
+ 
+ ## Integration
 
 - Can use output from `/ai-eng/research` via `--from-research`
 - Feeds into `/ai-eng/plan` for implementation planning

package/dist/.opencode/command/ai-eng/work.md

@@ -573,29 +573,14 @@ This will:
 ## Success Criteria
 
 A work session is successful when:
-- ✅ All tasks completed
-- ✅ All quality gates passed
-- ✅ All tests passing
-- ✅ Build succeeds
-- ✅ PR created and reviewed
-- ✅ Code merged to main
-
-## Execution
-
-Execute a plan using:
-
-```bash
-bun run scripts/run-command.ts work "$ARGUMENTS" [options]
-```
-
-For example:
-- `bun run scripts/run-command.ts work "specs/auth/plan.yaml" --verbose`
-- `bun run scripts/run-command.ts work --from-plan=plans/auth.yaml --continue`
-- `bun run scripts/run-command.ts work "FEAT-001" --dry-run`
-
-After completing implementation, rate your confidence in code quality and completeness (0.0-1.0). Identify any uncertainties about test coverage, areas where quality gates may have been too lenient, or implementation details that may need follow-up. Note any technical debt or known issues that should be tracked.
-
-## Integration
+ - ✅ All tasks completed
+ - ✅ All quality gates passed
+ - ✅ All tests passing
+ - ✅ Build succeeds
+ - ✅ PR created and reviewed
+ - ✅ Code merged to main
+ 
+ ## Integration
 
 - Reads from `/ai-eng/plan` output (plan.yaml)
 - Validates against `/ai-eng/specify` output (spec.md)

package/dist/index.js

@@ -68,8 +68,7 @@ var AiEngSystem = async ({
   const targetDir = findInstallationTarget(directory);
   if (!targetDir) {
     return {
-      config: async (input) => {
-      }
+      config: async (input) => {}
     };
   }
   const isFirstRun = !fs.existsSync(path.join(targetDir, "command", "ai-eng"));
@@ -82,8 +81,7 @@ var AiEngSystem = async ({
     console.error(`[ai-eng-system] Installation warning: ${error instanceof Error ? error.message : String(error)}`);
   }
   return {
-    config: async (input) => {
-    }
+    config: async (input) => {}
   };
 };
 var src_default = AiEngSystem;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-eng-system",
-  "version": "0.2.1",
+  "version": "0.2.3",
   "description": "Compounding engineering system for Claude Code and OpenCode. Shared agents, commands, skills, and plugin development tools.",
   "type": "module",
   "main": "./dist/index.js",
@@ -65,8 +65,6 @@
   ],
   "devDependencies": {
     "@biomejs/biome": "^1.9.4",
-    "@types/commander": "^2.12.0",
-    "@types/glob": "^8.1.0",
     "bun-types": "latest",
     "husky": "^9.1.7",
     "standard-version": "^9.5.0",