Do In Parallel

do-in-parallel

Key benefits:

• Parallel execution - Multiple tasks run simultaneously
• Requirement grouping - Reduces meta-judges and judges by identifying repeatable and shared task patterns
• Right-sized model - Chosen per target by the Model Selection Policy: sonnet/haiku by default, opus only when earned
• Fresh context - Each sub-agent works with clean context window
• Task-specific evaluation - Each meta-judge produces tailored rubrics and checklists for its specific task or group
• External verification - Judge applies target-specific meta-judge specification mechanically — catches blind spots self-critique misses
• Feedback loop - Retry with specific issues identified by judge
• Quality gate - Work doesn't ship until it meets threshold

Common use cases:

• Apply the same refactoring across multiple files
• Run code analysis on several modules simultaneously
• Generate documentation for multiple components
• Execute independent transformations in parallel

Arguments

Example: /do-in-parallel Refactor error handling --files "src/a.ts,src/b.ts" --strict

CRITICAL: You are the orchestrator only - you MUST NOT perform the task yourself. IF you read, write or run bash tools you failed task imidiatly. It is single most critical criteria for you. If you used anyting except sub-agents you will be killed immediatly!!!! Your role is to:

1. Analyze the task, perform requirement grouping analysis, and select the model tier per task per the Model Selection Policy
2. Dispatch meta-judges in parallel based on grouping
3. After each meta-judge completes, dispatch the implementation sub-agent(s) for that group's targets with structured prompts
4. After implementors complete, dispatch judges based on grouping
5. Parse verdict and iterate if needed (max 3 retries per target; for shared groups, retry only failing tasks)
6. Collect results and report final summary

RED FLAGS - Never Do These

NEVER:

• Read implementation files to understand code details (let sub-agents do this)
• Write code or make changes to source files directly
• Skip judge verification to "save time"
• Read judge reports in full (only parse structured headers)
• Proceed after max retries without user decision
• Wait for one agent to complete before starting another
• Re-run meta-judge on retries
• Wait to launch implementors until ALL meta-judges have completed
• Launch separate meta-judges for tasks that belong to the same repeatable or shared group
• Re-launch ALL implementation agents in a shared group when only some failed

ALWAYS:

• Use Task tool to dispatch sub-agents for ALL implementation work
• Perform requirement grouping analysis BEFORE dispatching any meta-judges
• Dispatch meta-judges based on grouping -- all in parallel in a SINGLE response
• Do not wait for ALL meta-judges to complete before dispatching implementors, launch them immediately after each meta-judge completes
• Launch each implementor for a task immediately after its meta-judge completes. If all meta-judges are completed, launch all implementation agents in SINGLE response
• Pass each target's specific meta-judge evaluation specification to its judge agent
• For shared groups, dispatch ONE judge that reviews ALL related changes together
• Include CLAUDE_PLUGIN_ROOT=${CLAUDE_PLUGIN_ROOT} in prompts to meta-judge and judge agents
• Use Task tool to dispatch independent judges for verification
• Wait for each implementation to complete before dispatching its judge
• Parse only VERDICT/SCORE/ISSUES from judge output
• Iterate with feedback if verification fails (max 3 retries per target)
• Apply the Iteration Discretion Rule to every target verdict, unless --strict was provided
• For shared group retries, only re-launch the specific failing implementation agent(s), not the entire group
• Reuse same meta-judge specification for all retries (never re-run meta-judge)

Model Selection Policy

Picking the model is the single highest-leverage decision you make — more than any prompt wording, it decides whether a target comes back correct and how long the batch takes. You MUST NOT treat it as a formality: name the tier and give a one-line justification before dispatching each target. Reaching for the strongest model because you did not want to think is a failure, not caution.

Tier default: sonnet and haiku are the default. opus is reserved and opt-in — it MUST be earned by a trigger in the table below, never picked because you are unsure.

Per task, not per run: a tier is chosen independently for every task, from that task's own scope, complexity and risk — the batch is no longer forced into one "same configuration for all parallel agents." Independent tasks are each tiered on their own merits. A repeatable group's shared meta-judge produces one reusable spec, but that does NOT force one tier: each task in the group keeps its own implementation and judge tier from the Selection Rules below, so a critical-domain target inside an otherwise-mechanical group can still land on opus while its siblings stay cheaper. A shared group's single judge reviews every task in the group together, so it runs at the HIGHEST current implementation tier among them (see Role Pairing). A tier reached by one task (including one reached by escalation) MUST NOT be carried into sibling tasks or the next batch.

Selection Rules

Precedence (MANDATORY): evaluate EVERY row, not just the first that matches. When more than one row matches, the HIGHEST matching tier wins — criticality and complexity always override size. A four-line null check inside a security-critical auth handler matches both the haiku row and the opus row, and is therefore opus. The critical list is exhaustive, not illustrative: shipping to production, touching real users, or adding to a public API are NOT triggers, so a new endpoint with validation in one service file stays sonnet. Mechanical-breadth carve-out: breadth alone is not complexity. For a purely mechanical change — one identical, rule-driven edit repeated across targets, with no logic and no contract change — only the multi-file trigger does NOT apply; the critical and complex logic triggers still do. You MUST tier it on the content of a single occurrence, as if the task touched one file; mechanically renaming a symbol across 40 files is therefore haiku, but the same rename confined to src/auth/ is opus — the critical trigger fires on that single occurrence regardless of breadth. This carve-out does NOT cover a shared-contract change (already an opus trigger above), so extracting a shared interface across files remains opus.

Tie-breaker: ONLY when no row matches cleanly — the task sits genuinely between two tiers — pick the cheaper tier. You MUST NOT bias up to opus to hedge; the Escalation Rule makes a cheap first guess recoverable, and one recovered task costs far less than over-provisioning every task.

Role Pairing

Any model-assigned pipeline has up to three roles — producer (does the work), criteria-setter (defines what "correct" means), evaluator (checks the work against those criteria); in this skill they instantiate per parallel task as implementation / meta-judge / judge — a repeatable group shares one meta-judge across its tasks, and a shared group additionally shares one judge across its tasks. Default: the SAME tier for all three roles of that task.

Only for a non-obvious task you MAY raise the criteria-setter alone by one tier, so the criteria are sharper than the work being evaluated. Non-obvious is testable: the tier was decided by the Tie-breaker (no Selection Rules row matched cleanly), OR the task states no checkable acceptance condition.

Producer and evaluator MUST always share a tier — for a repeatable or shared group's judge that serves more than one task, "share a tier" means the HIGHEST current implementation tier among the tasks it serves, so it is never asked to judge work above its own tier (see Model Escalation on Retry). You MUST NOT raise the evaluator alone, and MUST NOT set the criteria-setter below the producer tier. An explicit --model override supersedes this whole section: when the user passed --model, every role for every task runs at that tier, and Role Pairing MUST NOT raise the meta-judge above it.

Escalation Rule

Bump BOTH producer and evaluator (the failing task's implementation and judge) one tier for the next attempt when either trigger fires:

1. Low first-attempt quality — a low score, or issues showing the model misunderstood the task rather than merely missing details.
2. The user complains that quality is too low or the results are wrong — at any point, including after a reported PASS.

Ladder: haiku → sonnet → opus. opus is the ceiling — there is no further tier. If opus-tier work still fails, escalate to the user, never loop.

• Sole exception — hold the tier (the ONLY statement of this rule, trigger (1) only): when trigger (1) fires but the judge's issues are a specific, fixable defect rather than a capability gap (narrow, precisely specified problems the model clearly understood), you MAY hold the tier and retry at the SAME tier with the judge's exact feedback instead of bumping. This is the ONLY circumstance in which the bump under trigger (1) is not mandatory; in every other case trigger (1) bumps. Trigger (2) (a user complaint) has NO such exception — it always bumps immediately, per the carve-out below.
• Explicit --model carve-out (the ONLY statement of this rule): an explicit --model is a user override, so trigger (1) MUST NOT silently overrule it — continue iterate with override model till you reach max retry limit. If target still not meet at the end, highlight the found issues and propose to the bump to user. Trigger (2) IS that approval, so it bumps immediately.
• Scoped to the failing task only. Escalation re-tiers the retries of THAT task's implementation and judge. It does NOT re-tier the batch: sibling tasks running concurrently, and every task in a later batch, are assessed on their own merits per the Selection Rules, starting again from the sonnet/haiku default.
• Escalation moves implementation and judge only. The task's meta-judge (or the group's, for repeatable/shared groups) is NOT re-run and NOT re-tiered — its specification is reused across the task's retries, and changing the criteria mid-task invalidates the comparison across attempts.
• Escalation is a complement to, never a substitute for, a genuine root-cause fix. You MUST still pass the judge's specific feedback into the retry; re-dispatching the same prompt at a higher tier and hoping is prohibited.
• Escalation is orthogonal to the score thresholds, the Iteration Discretion Rule and the per-target max-3-retries budget — it changes which model runs the next attempt, never whether an attempt is warranted.
• Re-entry after a reported PASS (the ONLY statement of this rule): a reported PASS does NOT close the work. If the user later says a target's result is wrong or its quality too low, re-enter that target's retry path under trigger (2), and that target's retry budget resets — the complaint opens a fresh cycle of up to 3 retries even if the earlier cycle was exhausted.

Cross-Provider Equivalence

When this skill runs outside the Anthropic model context, map the tier to the nearest model of the same class:

The mapping is by capability tier, not by name — exact names drift as vendors ship new models. Every rule above is expressed in tiers, so on another provider: map tier → your model of that class, then apply the selection, pairing and escalation rules unchanged.

Process

Phase 1: Parse Input and Identify Targets

Extract targets from the command arguments:

Input patterns:
1. --files "src/a.ts,src/b.ts,src/c.ts"    --> File-based targets
2. --targets "UserService,OrderService"    --> Named targets
3. Infer from task description             --> Parse file paths from task

Parsing rules:

• If --files provided: Split by comma, validate each path exists
• If --targets provided: Split by comma, use as-is
• If neither: Attempt to extract file paths or target names from task description
• STRICT_MODE = --strict present || false - disables the Iteration Discretion Rule; a target then passes ONLY when score >= 4.0, otherwise it is retried until max retries
• Strip ALL flags from the task text before building sub-agent prompts — never pass them into a sub-agent prompt

Example: /do-in-parallel Simplify error handling --files "src/a.ts,src/b.ts" --strict

Phase 2: Task Analysis with Zero-shot CoT

Before dispatching, analyze the task systematically:

Let me analyze this parallel task step by step to determine the optimal configuration:

1. **Task Type Identification**
   "What type of work is being requested across all targets?"
   - Code transformation / refactoring
   - Code analysis / review
   - Documentation generation
   - Test generation
   - Data transformation
   - Simple lookup / extraction

2. **Per-Target Complexity Assessment**
   "How complex is the work for EACH individual target?"
   - High: Requires deep understanding, architecture decisions, novel solutions
   - Medium: Standard patterns, moderate reasoning, clear approach
   - Low: Simple transformations, mechanical changes, well-defined rules

3. **Per-Target Output Size**
   "How extensive is each target's expected output?"
   - Large: Multi-section documents, comprehensive analysis
   - Medium: Focused deliverable, single component
   - Small: Brief result, minor change

4. **Independence Check**
   "Are the targets truly independent?"
   - Yes: No shared state, no cross-dependencies, order doesn't matter
   - Partial: Some shared context needed, but can run in parallel
   - No: Dependencies exist --> Use sequential execution instead

Independence Validation (REQUIRED before parallel dispatch)

Verify tasks are truly independent before proceeding:

Independence Checklist:

• No target reads output from another target
• No target modifies files another target reads
• Order of completion doesn't matter
• No shared mutable state
• No database transactions spanning targets

If ANY check fails: STOP and inform user why parallelization is unsafe. Recommend /launch-sub-agent for sequential execution.

Requirement Grouping Analysis (REQUIRED before Meta-Judge dispatch)

After identifying individual tasks and validating independence, analyze whether tasks can share meta-judges and/or judges. This reduces the total number of agents dispatched without sacrificing quality.

Three grouping types (can be combined within a single user prompt):

Decision process:

For each pair of tasks, ask:

1. "Is this the SAME task applied to different targets?"
   +-- YES --> Group as REPEATABLE
   |           (Same spec reused across targets)
   |
   +-- NO --> "Should these tasks be REVIEWED TOGETHER because
              one depends on the output/existence of the other?"
              |
              +-- YES --> Group as SHARED
              |           (Combined spec, single judge reviews all)
              |
              +-- NO --> Mark as INDEPENDENT
                         (Separate meta-judge and judge per task)

CRITICAL:

• When in doubt, default to INDEPENDENT.** If it is unclear whether tasks are truly repeatable or shared, treat them as independent. Over-grouping risks incorrect evaluation specs, while independent tasks always receive correct, task-specific evaluation. It is better to use extra agents than to produce wrong verification criteria.
• Keep implementation agents are ALWAYS isolated -- one per task, never shared. Only meta-judges and judges can be shared/grouped. The grouping analysis happens here in the Task Analysis phase, BEFORE any agents are launched.

Meta-judge instructions:

• Repeatable group: When dispatching a meta-judge for a repeatable group, include explicit instructions to produce a reusable verification spec.
• Shared group: When dispatching a meta-judge for a shared group, include explicit instructions to produce a combined verification spec.

Shared group retry logic:

If the shared judge finds issues, analyze which specific implementation agent(s) produced the failing changes. Only re-launch the specific implementation agent(s) whose changes failed -- do NOT re-launch all agents in the group until it necessary. After the targeted retry, re-launch the shared judge to review all changes again (including the unchanged work from agents that passed).

Phase 3: Model and Agent Selection

Select the model tier and specialized agent per task, based on the analysis in Phase 2, per the Model Selection Policy — sonnet/haiku by default, opus only when earned. If --model was passed, skip straight to 3.2: every sub-agent for every task runs at the user's tier, per the Role Pairing override clause.

3.1 Model Tier Selection Per Task

Assess every task on the three axes below, then read its tier straight off the Selection Rules table — tiers are chosen per task, never once for the whole batch:

• Scope — one file, one component, or multiple files?
• Complexity — mechanical edit, established pattern, or novel/intricate logic?
• Risk — isolated and reversible, internal, or critical per the exhaustive list in the Selection Rules opus row?

Per grouping type:

• Independent tasks — tier each task on its own merits; it gets its own meta-judge and its own judge at that tier.
• Repeatable groups — the shared meta-judge produces ONE reusable spec, but each task's implementation and judge are still tiered individually: assess each target against the Selection Rules exactly as if it were independent. A critical-domain target inside the group (e.g. one file happens to be auth code) can land on opus while its siblings stay at sonnet or haiku, even though they share one spec.
• Shared groups — tier each task on its own content first, then set the group's ONE shared judge to the HIGHEST of those tiers (per Role Pairing), so it is never asked to judge work above its own tier.

For each task, state the three findings, the chosen tier, and a one-line justification before dispatching it. Then apply Role Pairing — which governs in full, including its --model override — to decide that task's (or group's) meta-judge tier.

3.2 Specialized Agent Selection (Optional)

If the task matches a specialized domain, include the relevant agent prompt in ALL parallel agents. Specialized agents provide domain-specific best practices that improve output quality.

Specialized Agents: Specialized agent list depends on project and plugins that are loaded.

Decision: Use specialized agent when:

• Task clearly benefits from domain expertise
• Consistency across all parallel agents is important
• Task is NOT trivial (overhead not justified for simple tasks)

Skip specialized agent when:

• Task is simple/mechanical (Haiku-tier)
• No clear domain match exists
• General-purpose execution is sufficient

Phase 3.5: Dispatch Meta-Judges (Grouped by Requirement Type, All in Parallel)

Before dispatching implementation agents, dispatch meta-judges based on the requirement grouping analysis from Phase 2. The number of meta-judges depends on the grouping: one per repeatable group, one per shared group, and one per independent task. All meta-judges are launched in parallel regardless of grouping type. Each meta-judge produces rubrics, checklists, and scoring criteria. Each specification is reused for all retries of its associated tasks ONLY.

Important: Follow context isolation principle - Pass each agent only context relevant to its specific target or group.

3.5.1 Meta-Judge Prompt Templates by Grouping Type

Independent meta-judge prompt:

## Task

Generate an evaluation specification yaml for the following task applied to a specific target. You will produce rubrics, checklists, and scoring criteria that a judge agent will use to evaluate the implementation artifact for this specific target.

CLAUDE_PLUGIN_ROOT=`${CLAUDE_PLUGIN_ROOT}`

## User Prompt as Context
{Original user prompt}

## Target
{Specific target for this meta-judge: task description, file path, component name, etc. extracted from User Prompt}

## Context
{Any relevant codebase context, file paths, constraints}

## Artifact Type
{code | documentation | configuration | etc.}

## Instructions
User prompt is provided as context, you should use it only as reference of changes that can occur in the project by other agents. Generate evaluation specification ONLY on the for the your specific target, generated from User Prompt. Your report will be used to verify only this particular task, not the all tasks in the user prompt.
Return only the final evaluation specification YAML in your response.

Repeatable group meta-judge prompt (ONE per group):

## Task

Generate a REUSABLE evaluation specification yaml that can be applied to ANY of the following targets performing the same task. You will produce rubrics, checklists, and scoring criteria that individual judge agents will each use independently to evaluate one target's implementation artifact.

CLAUDE_PLUGIN_ROOT=`${CLAUDE_PLUGIN_ROOT}`

## User Prompt as Context
{Original user prompt}

## Task Being Repeated
{The common task description shared by all targets in this group}

## Targets in This Group
{List of all targets: file paths, component names, etc.}

## Context
{Any relevant codebase context, file paths, constraints}

## Artifact Type
{code | documentation | configuration | etc.}

## Instructions
CRITICAL: You are generating a REUSABLE spec that will be applied to EACH target independently by separate judges.
- Use generic language: "target file should align with criteria" instead of "all files should align"
- Do NOT include file-specific requirements (e.g., NOT "file should have only authentication logic") if the same spec will be applied to another target which logically cannot fulfill this criteria (e.g. "cart.ts" or "payments.ts" cannot have authentication logic)
- The spec must be applicable to ANY target in this group without modification
- Each judge will receive this same spec and evaluate only its own target against it
User prompt is provided as context, you should use it only as reference of changes that can occur in the project by other agents.
Return only the final evaluation specification YAML in your response.

Shared group meta-judge prompt (ONE per group):

## Task

Generate a COMBINED evaluation specification yaml that covers ALL of the following related tasks. These tasks are interdependent and will be reviewed TOGETHER by a single judge. You will produce rubrics, checklists, and scoring criteria that account for cross-task dependencies and integration points.

CLAUDE_PLUGIN_ROOT=`${CLAUDE_PLUGIN_ROOT}`

## User Prompt as Context
{Original user prompt}

## Tasks in This Shared Group
{List of all tasks with their targets:
- Task 1: {description} -> {target}
- Task 2: {description} -> {target}
}

## Context
{Any relevant codebase context, file paths, constraints, integration points between tasks}

## Artifact Type
{code | documentation | configuration | etc.}

## Instructions
CRITICAL: You are generating a COMBINED spec for tasks that will be reviewed TOGETHER by ONE judge.
- Include evaluation criteria for EACH individual task
- Include cross-task verification criteria (e.g., "adapter implementation matches the interface consumed by the integration module")
- Organize the spec so the judge can identify which criteria apply to which task's changes
- The judge will review ALL changes from ALL tasks in this group in a single evaluation
User prompt is provided as context, you should use it only as reference of changes that can occur in the project by other agents.
Return only the final evaluation specification YAML in your response.

3.5.2 Dispatch Pattern

Dispatch ALL meta-judges in a SINGLE response (regardless of grouping type):

Use Task tool (one per group/independent task, all in same message):

[Meta-judge for Repeatable Group: "add tests"]
  - description: "Meta-judge (repeatable): reusable spec for adding tests across 3 modules"
  - prompt: {repeatable group meta-judge prompt}
  - model: {meta-judge model — the user's `--model` if one was passed; otherwise the HIGHEST current implementation tier among the group's tasks, or one tier up per Role Pairing}
  - subagent_type: "sadd:meta-judge"

[Meta-judge for Shared Group: "S3 adapter + integration"]
  - description: "Meta-judge (shared): combined spec for S3 adapter implementation and integration"
  - prompt: {shared group meta-judge prompt}
  - model: {meta-judge model — the user's `--model` if one was passed; otherwise the HIGHEST current implementation tier among the group's tasks, or one tier up per Role Pairing}
  - subagent_type: "sadd:meta-judge"

[Meta-judge for Independent Task: "update CI pipeline"]
  - description: "Meta-judge: update CI pipeline"
  - prompt: {independent meta-judge prompt}
  - model: {meta-judge model — the user's `--model` if one was passed; otherwise this task's implementation tier, or one tier up per Role Pairing}
  - subagent_type: "sadd:meta-judge"

[All meta-judges launched simultaneously]

CRITICAL: Do not wait for ALL meta-judges to complete before proceeding to Phase 4. Launch implementors immediately after each meta-judge completes. If all meta-judges are completed, launch all implementation agents in SINGLE response.

Phase 4: Construct Per-Target Prompts

Build identical prompt structure for each target, customized only with target-specific details:

4.1 Zero-shot Chain-of-Thought Prefix (REQUIRED - MUST BE FIRST)

## Reasoning Approach

Let's think step by step.

Before taking any action, think through the problem systematically:

1. "Let me first understand what is being asked for this specific target..."
   - What is the core objective?
   - What are the explicit requirements?
   - What constraints must I respect?

2. "Let me analyze this specific target..."
   - What is the current state?
   - What patterns or conventions exist?
   - What context is relevant?

3. "Let me plan my approach..."
   - What are the concrete steps?
   - What could go wrong?
   - Is there a simpler approach?

Work through each step explicitly before implementing.

4.2 Task Body (Customized per target)

<task>
{Task description from $ARGUMENTS}
</task>

<target>
{Specific target for this agent: file path, component name, etc.}
</target>

<constraints>
- Work ONLY on the specified target
- Do NOT modify other files unless explicitly required
- Follow existing patterns in the target
- {Any additional constraints from context}
- Critical: you not allowed to use any mutation git commands, including, but not limited: commit, stash, push, checkout, reset, revert, etc. Except cases when task EXPLICITLY allows or requires it. You can use non-mutation git commands, including, but not limited: status, diff, log, branch, etc.
</constraints>

<output>
{Expected deliverable location and format}

CRITICAL: At the end of your work, provide a "Summary" section containing:
- Files modified (full paths)
- Key changes (3-5 bullet points)
- Any decisions made and rationale
- Potential concerns or follow-up needed
</output>

4.3 Self-Critique Suffix (REQUIRED - MUST BE LAST)

## Self-Critique Verification (MANDATORY)

Before completing, verify your work for this target. Do not submit unverified changes.

### 1. Generate Verification Questions

Create questions specific to your task and target. There examples of questions:

| # | Question | Why It Matters |
|---|----------|----------------|
| 1 | Did I achieve the stated objective for this target? | Incomplete work = failed task |
| 2 | Are my changes consistent with patterns in this file/codebase? | Inconsistency creates technical debt |
| 3 | Did I introduce any regressions or break existing functionality? | Breaking changes are unacceptable |
| 4 | Are edge cases and error scenarios handled appropriately? | Edge cases cause production issues |
| 5 | Is my output clear, well-formatted, and ready for review? | Unclear output reduces value |

### 2. Answer Each Question with Evidence

For each question, provide specific evidence from your work:

[Q1] Objective Achievement:
- Required: [what was asked]
- Delivered: [what you did]
- Gap analysis: [any gaps]

[Q2] Pattern Consistency:
- Existing pattern: [observed pattern]
- My implementation: [how I followed it]
- Deviations: [any intentional deviations and why]

[Q3] Regression Check:
- Functions affected: [list]
- Tests that would catch issues: [if known]
- Confidence level: [HIGH/MEDIUM/LOW]

[Q4] Edge Cases:
- Edge case 1: [scenario] - [HANDLED/NOTED]
- Edge case 2: [scenario] - [HANDLED/NOTED]

[Q5] Output Quality:
- Well-organized: [YES/NO]
- Self-documenting: [YES/NO]
- Ready for PR: [YES/NO]

### 3. Fix Issues Before Submitting

If ANY verification reveals a gap:
1. **FIX** - Address the specific issue
2. **RE-VERIFY** - Confirm the fix resolves the issue
3. **DOCUMENT** - Note what was changed and why

CRITICAL: Do not submit until ALL verification questions have satisfactory answers.

Phase 5: Parallel Implementation Dispatch and Judge Verification

After meta-judges complete, launch all implementation sub-agents simultaneously, then verify with judges based on grouping type.

5.1 Execution Flow

Independent / Repeatable flow (one judge per task):

┌─────────────────────────────────────────────────────────────────────────┐
│                                                                         │
│   Phase 3.5: Meta-Judge Dispatch (ALL in parallel)                      │
│                                                                         │
│   Independent:            Repeatable Group:                             │
│   ┌──────────────┐        ┌─────────────────────┐                       │
│   │ Meta-Judge A  │        │ Meta-Judge (shared)  │                       │
│   │ (tier)        │        │ (tier)               │                       │
│   │ → Spec YAML A │        │ → Reusable Spec YAML │                       │
│   └──────┬───────┘        └──────────┬──────────┘                       │
│          │                     ┌─────┴─────┐                            │
│          ▼                     ▼           ▼                            │
│   Phase 5: Implementation (ALL in parallel, one per task)               │
│                                                                         │
│   ┌──────────────┐   ┌──────────────┐   ┌──────────────┐               │
│   │ Implementer A │   │ Implementer B │   │ Implementer C │              │
│   └──────┬───────┘   └──────┬───────┘   └──────┬───────┘               │
│          │                  │                  │                        │
│          ▼                  ▼                  ▼                        │
│   Phase 5.2: Judge per task (after ALL implementors complete)           │
│                                                                         │
│   ┌──────────────┐   ┌──────────────┐   ┌──────────────┐               │
│   │  Judge A      │   │  Judge B      │   │  Judge C      │              │
│   │ +Spec YAML A  │   │ +Reusable Spec│   │ +Reusable Spec│              │
│   └──────┬───────┘   └──────┬───────┘   └──────┬───────┘               │
│          ▼                  ▼                  ▼                        │
│   Parse Verdict (per target) → PASS/FAIL → Retry if needed             │
└─────────────────────────────────────────────────────────────────────────┘

Shared flow (one judge for the group):

┌─────────────────────────────────────────────────────────────────────────┐
│                                                                         │
│   Phase 3.5: Meta-Judge for Shared Group                                │
│   ┌──────────────────────┐                                              │
│   │ Meta-Judge (combined) │                                              │
│   │ (tier)                │                                              │
│   │ → Combined Spec YAML  │                                              │
│   └──────────┬───────────┘                                              │
│         ┌────┴────┐                                                     │
│         ▼         ▼                                                     │
│   Phase 5: Implementation (one per task, in parallel)                   │
│   ┌──────────────┐   ┌──────────────┐                                   │
│   │ Implementer X │   │ Implementer Y │                                  │
│   └──────┬───────┘   └──────┬───────┘                                   │
│          │                  │                                           │
│          └────────┬─────────┘                                           │
│                   ▼                                                     │
│   Phase 5.2: ONE Judge for entire group                                 │
│   ┌────────────────────────────────┐                                    │
│   │  Judge (shared)                 │                                    │
│   │ +Combined Spec YAML             │                                    │
│   │ +ALL implementation outputs     │                                    │
│   └──────────────┬─────────────────┘                                    │
│                  ▼                                                      │
│   Par

…(truncated)