antigravity-atomic-swarms 2.1.0 → 2.1.2

package/.agent/rules/02-sparc-specification/global-laws.md

@@ -1,92 +1,92 @@
-# Global Laws: 02-SPARC-Specification
-
-> **Applies To**: ALL phases in 02-SPARC-Specification workflow
-> **Enforcement**: MANDATORY - violations are critical failures
-
-## 🛑 SYSTEM OVERRIDE: CRITICAL GLOBAL LAWS
-
-### Law 1: The "No Silent Failures" Law
-**Rule**: NEVER write empty catch blocks. If a tool fails, STOP.
-**Enforcement**: Check every `try/catch`. If `catch(e) {}` exists -> REJECT.
-
-### Law 2: The "Clean Floor" Law
-**Rule**: Delete all temp files (`debug.json`, etc) before exiting.
-**Enforcement**: Use `Remove-Item` or `fs.unlink` on any temp artifact.
-
-### Law 3: The "Double-Tap" Law
-**Rule**: Verify EVERY file write by reading it back. Do not assume success.
-**Enforcement**: Call `view_file` immediately after `write_to_file`.
-
-### Law 4: The "Think-Tree" Law
-**Rule**: Before complex actions, visualize 3 paths. Choose the best.
-**Enforcement**: Use SCoT blocks (Law 5) for this visualization.
-
-### Law 5: The "SCoT Protocol" Law
-**Rule**: Before ANY decision with >1 viable path, you MUST output a `<scot>` block. Decisions without documented reasoning are invalid.
-**Enforcement**:
-```markdown
-<scot>
-**📍 SITUATION**: ...
-**🎯 GOAL**: ...
-**🔀 OPTIONS**: ...
-**🧪 ANALYSIS**: ...
-**✅ DECISION**: ...
-</scot>
-```
-
-### Law 6: The "Confidence Declaration" Law
-**Rule**: Every major decision MUST include a Confidence Score (0-100%) and Basis.
-**Enforcement**: Output CONFIDENCE: [Score]% and BASIS: [Reasoning] with every decision.
-
-### Law 7: The "Tool Validation" Law
-**Rule**: Before ANY tool invocation, verify schema in `memory/schemas/tool_registry.md`.
-**Enforcement**: Log intent and schema validation status.
-
-### Law 8: The "Escalation Protocol" Law
-**Rule**: If a decision triggers a T1, T2, or T3 event (see KNOWLEDGE_BASE/Escalation), you **MUST** pause and request human approval.
-**Enforcement**: Check `APPROVAL_LOG.md` before proceeding. If blocked, use `notify_user`.
-
-### Law 12: The "Autonomy Protocol" Law
-**Rule**: Adhere to `PROJECT_STATE.Config.Autonomy`.
-- **Manual**: Stop & Ask before EVERY file write/command. (User acts as Human-in-the-Loop).
-- **Semi-Guided**: Stop & Ask for Architectural Decisions (SCoT) or Critical Law warnings. Auto-proceed for routine.
-- **Fully Auto**: Auto-proceed unless AGI-level danger. Auto-Push.
-**Enforcement**: Check `PROJECT_STATE.md` at start of session.
-
-### Law 13: The "Evolution Imperative" Law
-**Rule**: If a tool fails 3+ times OR you identify a recurring anti-pattern, you **MUST** trigger the Evolution Protocol.
-**Enforcement**:
-1. Log to `memory/failure_log.md`.
-2. Check `PROJECT_STATE.EvolutionMode`.
-   - **Off**: Do nothing.
-   - **Semi/Full**: Consult `.agent/rules/EVOLUTION_PROTOCOL.md` to propose/execute a rule update.
-
----
-
-## System Constraints (Specific to Specification)
-
-### 1. The "Failure First" Ratio
-Happy Paths are easy. For every **1 Success Scenario**, you MUST define at least **3 Failure Scenarios** (e.g., "Card Declined", "Network Timeout"). Specs with only Happy Paths are **rejected**.
-
-### 2. The "Type Truth" Protocol
-Data structures must be defined in **TypeScript Interfaces**, not English. Don't say "list of users". Say: `Promise<Array<{id: string}>>`.
-
-### 3. The "Ambiguity Assassin"
-Bounds must be numeric. Define **Limits** for every input: Min/Max Length, Allowed Characters (Regex), Max File Size, Timeout Duration.
-
-### 4. The "Zombie State" Check
-Every state transition must have a defined **Reset/Exit** or **Error** state. If "Loading", define how to exit if the network dies.
-
-### 5. The "Security Mirror"
-Every Endpoint/Function spec must have a "Security Implications" section: "Auth Required?", "Input Sanitized?", "Rate Limited?", "PII Leaked?".
-
----
-
-## Violations & Consequences
-| Violation | Severity | Action |
-|-----------|----------|--------|
-| Missing Failures | HIGH | Reject Spec, Restart Phase 2 |
-| English Types | MEDIUM | Refactor to TS |
-| Missing SCoT | HIGH | Re-evaluate decision |
-| Tool Misuse | CRITICAL | Stop execution, Check Schema |
-| Unauthorized Escalation | CRITICAL | Rollback, Request Approval |
+# Global Laws: 02-SPARC-Specification
+
+> **Applies To**: ALL phases in 02-SPARC-Specification workflow
+> **Enforcement**: MANDATORY - violations are critical failures
+
+## 🛑 SYSTEM OVERRIDE: CRITICAL GLOBAL LAWS
+
+### Law 1: The "No Silent Failures" Law
+**Rule**: NEVER write empty catch blocks. If a tool fails, STOP.
+**Enforcement**: Check every `try/catch`. If `catch(e) {}` exists -> REJECT.
+
+### Law 2: The "Clean Floor" Law
+**Rule**: Delete all temp files (`debug.json`, etc) before exiting.
+**Enforcement**: Use `Remove-Item` or `fs.unlink` on any temp artifact.
+
+### Law 3: The "Double-Tap" Law
+**Rule**: Verify EVERY file write by reading it back. Do not assume success.
+**Enforcement**: Call `view_file` immediately after `write_to_file`.
+
+### Law 4: The "Think-Tree" Law
+**Rule**: Before complex actions, visualize 3 paths. Choose the best.
+**Enforcement**: Use SCoT blocks (Law 5) for this visualization.
+
+### Law 5: The "SCoT Protocol" Law
+**Rule**: Before ANY decision with >1 viable path, you MUST output a `<scot>` block. Decisions without documented reasoning are invalid.
+**Enforcement**:
+```markdown
+<scot>
+**📍 SITUATION**: ...
+**🎯 GOAL**: ...
+**🔀 OPTIONS**: ...
+**🧪 ANALYSIS**: ...
+**✅ DECISION**: ...
+</scot>
+```
+
+### Law 6: The "Confidence Declaration" Law
+**Rule**: Every major decision MUST include a Confidence Score (0-100%) and Basis.
+**Enforcement**: Output CONFIDENCE: [Score]% and BASIS: [Reasoning] with every decision.
+
+### Law 7: The "Tool Validation" Law
+**Rule**: Before ANY tool invocation, verify schema in `memory/schemas/tool_registry.md`.
+**Enforcement**: Log intent and schema validation status.
+
+### Law 8: The "Escalation Protocol" Law
+**Rule**: If a decision triggers a T1, T2, or T3 event (see KNOWLEDGE_BASE/Escalation), you **MUST** pause and request human approval.
+**Enforcement**: Check `APPROVAL_LOG.md` before proceeding. If blocked, use `notify_user`.
+
+### Law 12: The "Autonomy Protocol" Law
+**Rule**: Adhere to `PROJECT_STATE.Config.Autonomy`.
+- **Manual**: Stop & Ask before EVERY file write/command. (User acts as Human-in-the-Loop).
+- **Semi-Guided**: Stop & Ask for Architectural Decisions (SCoT) or Critical Law warnings. Auto-proceed for routine.
+- **Fully Auto**: Auto-proceed unless AGI-level danger. Auto-Push.
+**Enforcement**: Check `PROJECT_STATE.md` at start of session.
+
+### Law 13: The "Evolution Imperative" Law
+**Rule**: If a tool fails 3+ times OR you identify a recurring anti-pattern, you **MUST** trigger the Evolution Protocol.
+**Enforcement**:
+1. Log to `memory/failure_log.md`.
+2. Check `PROJECT_STATE.EvolutionMode`.
+   - **Off**: Do nothing.
+   - **Semi/Full**: Consult `.agent/rules/EVOLUTION_PROTOCOL.md` to propose/execute a rule update.
+
+---
+
+## System Constraints (Specific to Specification)
+
+### 1. The "Failure First" Ratio
+Happy Paths are easy. For every **1 Success Scenario**, you MUST define at least **3 Failure Scenarios** (e.g., "Card Declined", "Network Timeout"). Specs with only Happy Paths are **rejected**.
+
+### 2. The "Type Truth" Protocol
+Data structures must be defined in **TypeScript Interfaces**, not English. Don't say "list of users". Say: `Promise<Array<{id: string}>>`.
+
+### 3. The "Ambiguity Assassin"
+Bounds must be numeric. Define **Limits** for every input: Min/Max Length, Allowed Characters (Regex), Max File Size, Timeout Duration.
+
+### 4. The "Zombie State" Check
+Every state transition must have a defined **Reset/Exit** or **Error** state. If "Loading", define how to exit if the network dies.
+
+### 5. The "Security Mirror"
+Every Endpoint/Function spec must have a "Security Implications" section: "Auth Required?", "Input Sanitized?", "Rate Limited?", "PII Leaked?".
+
+---
+
+## Violations & Consequences
+| Violation | Severity | Action |
+|-----------|----------|--------|
+| Missing Failures | HIGH | Reject Spec, Restart Phase 2 |
+| English Types | MEDIUM | Refactor to TS |
+| Missing SCoT | HIGH | Re-evaluate decision |
+| Tool Misuse | CRITICAL | Stop execution, Check Schema |
+| Unauthorized Escalation | CRITICAL | Rollback, Request Approval |

package/.agent/rules/02-sparc-specification/phase--0.6-constitutional-compliance.md

@@ -1,50 +1,50 @@
-# Phase -0.6: Constitutional Compliance Check
-
-> **Parent Workflow**: 02-SPARC-Specification
-> **Phase Number**: -0.6
-> **Objective**: Ensure alignment with supreme laws before execution.
-
----
-
-## Prerequisites
-**MUST be completed before this phase**:
-- ✅ None (First Phase)
-
-**MUST have access to**:
-- `memory/CONSTITUTION.md`
-
----
-
-## Instructions
-
-### Step 1: Read Constitution
-**Action**: Load the Constitution file.
-**Exact Prompt**:
-```
-> 📖 READ AND EXECUTE: `memory/CONSTITUTION.md`
-```
-
-### Step 2: Self-Audit
-**Action**: Ask yourself:
-1. "Does my intended next action violate any Article?"
-2. "Am I currently in a loop state (Breaker 2.1)?"
-
-### Step 3: Circuit Breaker Logic
-**Action**: Evaluate safety.
-- **If Violation**:
-    - STOP immediately.
-    - Log to `failure_log.md`: "COMBAT BREAKER TRIPPED: [Article]"
-    - Emergency Exit to Notify User.
-- **If Safe**: Proceed to Phase -0.5 (or next available phase).
-
----
-
-## Exit Criteria
-**This phase is complete when ALL of the following are true**:
-- ✅ Constitution has been read.
-- ✅ Self-Audit confirmed no violations.
-
----
-
-## Related Phases
-**Next Phase**: Phase 0: The Ancestral Audit
+# Phase -0.6: Constitutional Compliance Check
+
+> **Parent Workflow**: 02-SPARC-Specification
+> **Phase Number**: -0.6
+> **Objective**: Ensure alignment with supreme laws before execution.
+
+---
+
+## Prerequisites
+**MUST be completed before this phase**:
+- ✅ None (First Phase)
+
+**MUST have access to**:
+- `memory/CONSTITUTION.md`
+
+---
+
+## Instructions
+
+### Step 1: Read Constitution
+**Action**: Load the Constitution file.
+**Exact Prompt**:
+```
+> 📖 READ AND EXECUTE: `memory/CONSTITUTION.md`
+```
+
+### Step 2: Self-Audit
+**Action**: Ask yourself:
+1. "Does my intended next action violate any Article?"
+2. "Am I currently in a loop state (Breaker 2.1)?"
+
+### Step 3: Circuit Breaker Logic
+**Action**: Evaluate safety.
+- **If Violation**:
+    - STOP immediately.
+    - Log to `failure_log.md`: "COMBAT BREAKER TRIPPED: [Article]"
+    - Emergency Exit to Notify User.
+- **If Safe**: Proceed to Phase -0.5 (or next available phase).
+
+---
+
+## Exit Criteria
+**This phase is complete when ALL of the following are true**:
+- ✅ Constitution has been read.
+- ✅ Self-Audit confirmed no violations.
+
+---
+
+## Related Phases
+**Next Phase**: Phase 0: The Ancestral Audit

package/.agent/rules/02-sparc-specification/phase--1-session-entry.md

@@ -1,89 +1,89 @@
-# Phase -1: Session Entry Protocol (MANDATORY - CANNOT BE SKIPPED)
-
-> **Parent Workflow**: 02-SPARC-Specification
-> **Phase Number**: -1
-> **Objective**: Establish perfect situational awareness and apply lessons learned.
-
----
-
-## Prerequisites
-**MUST be completed before this phase**:
-- [x] None (First Phase)
-
-**MUST have access to**:
-- `memory/PROJECT_STATE.md`
-- `memory/SESSION_HANDOFF.md`
-- `memory/EPISODIC_MEMORY.md`
-- `memory/failure_log.md`
-- `memory/KNOWLEDGE_BASE.md`
-
----
-
-## Instructions
-
-### Step 1: Full Context Absorption (State Awareness)
-**Action**: Read ALL of the following files in order:
-1. `memory/PROJECT_STATE.md` (canonical truth)
-2. `memory/SESSION_HANDOFF.md` (previous session's exit state)
-3. `memory/failure_log.md` (historical lessons)
-4. `memory/KNOWLEDGE_BASE.md` (patterns and anti-patterns)
-5. `memory/MEMORY_STREAM.md` (volatile context)
-
-### Step 2: State Awareness Report
-**Action**: Output the following report to the User (ALWAYS visible, cannot be suppressed):
-
-```
-+----------------------------------- STATE AWARENESS REPORT -----------------------------------+
-| PROJECT: [Name from PROJECT_STATE]                                                           |
-| LAST WORKFLOW: [Workflow that last touched PROJECT_STATE]                                    |
-| CURRENT PHASE: [Exact status e.g., "(R) Researching"]                                        |
-| FEATURE IN FOCUS: [Which feature is currently being worked]                                  |
-|                                                                                              |
-| COMPLETED PHASES:                                                                            |
-|   [x] 00-Solicitation: [Date] - [Summary]                                                    |
-|   [x] 01-Research: [Date] - [Summary]                                                        |
-|   [.] 02-Specification: IN PROGRESS                                                          |
-|                                                                                              |
-| PENDING ACTIONS:                                                                             |
-|   1. [Exact next action from last workflow exit]                                             |
-|                                                                                              |
-| LEARNED CONSTRAINTS (from failure_log):                                                      |
-|   (!) [Rule 1 from ancestral lessons]                                                        |
-|   (!) [Rule 2 from ancestral lessons]                                                        |
-+----------------------------------------------------------------------------------------------+
-```
-
-### Step 3: Episodic Memory Injection (Reflexion Pattern)
-**Action**:
-1.  **Episode Search**: Scan `memory/EPISODIC_MEMORY.md` for matching episodes.
-2.  **Pattern Match**:
-    - If match found: Load `Lesson Learned` and increment `Reference Count`. Log: "(Ref) Applying lesson from [EPI-XXX]: [Lesson]".
-    - If no match: Proceed.
-3.  **Session Episode Creation**: Create draft episode for this session.
-
-### Step 4: User Verification Gate
-**Action**: Ask: "I have absorbed all context. Does this summary match your understanding?"
-- **Fully Auto**: Proceed after 2s pause.
-- **Semi/Manual**: Wait for explicit "Yes".
-
-### Step 5: Memory Decay Check
-**Action**: Once per session, scan for episodes > 30 days old with Reference Count < 3. Move to `memory/archive/`.
-
----
-
-
-
-### Step 0.5: Health Check
-**Action**: Read memory/HEALTH_DASHBOARD.md.
-**Action**: Note any CRITICAL issues before proceeding.
-
-## Exit Criteria
-**This phase is complete when ALL of the following are true**:
-- [x] State Awareness Report displayed.
-- [x] Relevant episodic memory loaded.
-- [x] User (or timer) approved proceeds.
-
----
-
-## Related Phases
-**Next Phase**: Phase -0.6: Constitutional Compliance Check
+# Phase -1: Session Entry Protocol (MANDATORY - CANNOT BE SKIPPED)
+
+> **Parent Workflow**: 02-SPARC-Specification
+> **Phase Number**: -1
+> **Objective**: Establish perfect situational awareness and apply lessons learned.
+
+---
+
+## Prerequisites
+**MUST be completed before this phase**:
+- [x] None (First Phase)
+
+**MUST have access to**:
+- `memory/PROJECT_STATE.md`
+- `memory/SESSION_HANDOFF.md`
+- `memory/EPISODIC_MEMORY.md`
+- `memory/failure_log.md`
+- `memory/KNOWLEDGE_BASE.md`
+
+---
+
+## Instructions
+
+### Step 1: Full Context Absorption (State Awareness)
+**Action**: Read ALL of the following files in order:
+1. `memory/PROJECT_STATE.md` (canonical truth)
+2. `memory/SESSION_HANDOFF.md` (previous session's exit state)
+3. `memory/failure_log.md` (historical lessons)
+4. `memory/KNOWLEDGE_BASE.md` (patterns and anti-patterns)
+5. `memory/MEMORY_STREAM.md` (volatile context)
+
+### Step 2: State Awareness Report
+**Action**: Output the following report to the User (ALWAYS visible, cannot be suppressed):
+
+```
++----------------------------------- STATE AWARENESS REPORT -----------------------------------+
+| PROJECT: [Name from PROJECT_STATE]                                                           |
+| LAST WORKFLOW: [Workflow that last touched PROJECT_STATE]                                    |
+| CURRENT PHASE: [Exact status e.g., "(R) Researching"]                                        |
+| FEATURE IN FOCUS: [Which feature is currently being worked]                                  |
+|                                                                                              |
+| COMPLETED PHASES:                                                                            |
+|   [x] 00-Solicitation: [Date] - [Summary]                                                    |
+|   [x] 01-Research: [Date] - [Summary]                                                        |
+|   [.] 02-Specification: IN PROGRESS                                                          |
+|                                                                                              |
+| PENDING ACTIONS:                                                                             |
+|   1. [Exact next action from last workflow exit]                                             |
+|                                                                                              |
+| LEARNED CONSTRAINTS (from failure_log):                                                      |
+|   (!) [Rule 1 from ancestral lessons]                                                        |
+|   (!) [Rule 2 from ancestral lessons]                                                        |
++----------------------------------------------------------------------------------------------+
+```
+
+### Step 3: Episodic Memory Injection (Reflexion Pattern)
+**Action**:
+1.  **Episode Search**: Scan `memory/EPISODIC_MEMORY.md` for matching episodes.
+2.  **Pattern Match**:
+    - If match found: Load `Lesson Learned` and increment `Reference Count`. Log: "(Ref) Applying lesson from [EPI-XXX]: [Lesson]".
+    - If no match: Proceed.
+3.  **Session Episode Creation**: Create draft episode for this session.
+
+### Step 4: User Verification Gate
+**Action**: Ask: "I have absorbed all context. Does this summary match your understanding?"
+- **Fully Auto**: Proceed after 2s pause.
+- **Semi/Manual**: Wait for explicit "Yes".
+
+### Step 5: Memory Decay Check
+**Action**: Once per session, scan for episodes > 30 days old with Reference Count < 3. Move to `memory/archive/`.
+
+---
+
+
+
+### Step 0.5: Health Check
+**Action**: Read memory/HEALTH_DASHBOARD.md.
+**Action**: Note any CRITICAL issues before proceeding.
+
+## Exit Criteria
+**This phase is complete when ALL of the following are true**:
+- [x] State Awareness Report displayed.
+- [x] Relevant episodic memory loaded.
+- [x] User (or timer) approved proceeds.
+
+---
+
+## Related Phases
+**Next Phase**: Phase -0.6: Constitutional Compliance Check

package/.agent/rules/02-sparc-specification/phase-0-ancestral-audit.md

@@ -1,44 +1,44 @@
-# Phase 0: The Ancestral Audit (Wisdom Access)
-
-> **Parent Workflow**: 02-SPARC-Specification
-> **Phase Number**: 0
-> **Objective**: Avoid repeating the "Lazy Spec" mistakes of the past.
-
----
-
-## Prerequisites
-**MUST be completed before this phase**:
-- ✅ Phase -0.6: Constitutional Compliance Check
-
-**MUST have access to**:
-- `memory/failure_log.md`
-- `memory/PROJECT_STATE.md`
-
----
-
-## Instructions
-
-### Step 1: Wisdom Retrieval
-**Action**: Read `memory/failure_log.md` (if exists) and `memory/PROJECT_STATE.md` (history).
-**Prompt**: "Review the `failure_log.md`. Did previous features fail due to bad specs (e.g., 'Undefined Types', 'Missed Edge Cases')? List 3 constrained rules to prevent recurrence."
-
-### Step 2: Mode Check
-- **If 'HOTFIX_REQUEST' in `MEMORY_STREAM`**: Enter **Hotfix Mode** (Spec ONLY the fix -> `03-london-tdd-builder`).
-- **Else**: Proceed to Phase 1 (Normal Mode).
-
-### Step 3: Phase Checklist
-**Action**: Open `.agent/checklists/02-sparc-specification/phase-0-ancestral-audit-checklist.md`.
-**Instruction**: Verify every item is checked.
-
----
-
-## Exit Criteria
-**This phase is complete when ALL of the following are true**:
-- ✅ Previous failures reviewed.
-- ✅ Mode selected.
-- ✅ Checklist verified.
-
----
-
-## Related Phases
-**Next Phase**: Phase 1: Job Hunt & Context Absorption
+# Phase 0: The Ancestral Audit (Wisdom Access)
+
+> **Parent Workflow**: 02-SPARC-Specification
+> **Phase Number**: 0
+> **Objective**: Avoid repeating the "Lazy Spec" mistakes of the past.
+
+---
+
+## Prerequisites
+**MUST be completed before this phase**:
+- ✅ Phase -0.6: Constitutional Compliance Check
+
+**MUST have access to**:
+- `memory/failure_log.md`
+- `memory/PROJECT_STATE.md`
+
+---
+
+## Instructions
+
+### Step 1: Wisdom Retrieval
+**Action**: Read `memory/failure_log.md` (if exists) and `memory/PROJECT_STATE.md` (history).
+**Prompt**: "Review the `failure_log.md`. Did previous features fail due to bad specs (e.g., 'Undefined Types', 'Missed Edge Cases')? List 3 constrained rules to prevent recurrence."
+
+### Step 2: Mode Check
+- **If 'HOTFIX_REQUEST' in `MEMORY_STREAM`**: Enter **Hotfix Mode** (Spec ONLY the fix -> `03-london-tdd-builder`).
+- **Else**: Proceed to Phase 1 (Normal Mode).
+
+### Step 3: Phase Checklist
+**Action**: Open `.agent/checklists/02-sparc-specification/phase-0-ancestral-audit-checklist.md`.
+**Instruction**: Verify every item is checked.
+
+---
+
+## Exit Criteria
+**This phase is complete when ALL of the following are true**:
+- ✅ Previous failures reviewed.
+- ✅ Mode selected.
+- ✅ Checklist verified.
+
+---
+
+## Related Phases
+**Next Phase**: Phase 1: Job Hunt & Context Absorption

package/.agent/rules/02-sparc-specification/phase-0.5-meta-strategy.md

@@ -1,54 +1,54 @@
-# Phase 0.5: Meta-Cognitive Strategy Selection
-> **Parent Workflow**: 02-SPARC-Specification
-> **Phase Number**: 0.5
-> **Objective**: Adapt workflow behavior based on past performance.
-
----
-
-## Prerequisites
-**MUST be completed before this phase**:
-- ✅ Phase 0: Ancestral Audit
-
----
-
-## Instructions
-
-### Step 1: Read Performance Metrics
-**Action**: Read `memory/PERFORMANCE_LOG.md`.
-
-### Step 2: Pattern Analysis
-**Action**: Calculate the following from the last 3 entries:
-1.  **Avg Loops**: (Sum of loops) / 3
-2.  **Avg Mirror Score**: (Sum of scores) / 3
-
-### Step 3: Strategy Selection
-**Action**: Select one strategy strategy based on the data:
-
-| Condition | Strategy | Behavior |
-|-----------|----------|----------|
-| Avg loops > 5 | **VERBOSE** | Extra logging, more checkpoints |
-| Mirror Score < 8 | **CAUTIOUS** | Self-critique after every action |
-| BMO/Test Failures > 1 | **PARANOID** | Double-verify all assertions |
-| All metrics good | **STREAMLINED** | Skip optional checkpoints |
-| No Data | **STANDARD** | Default behavior |
-
-### Step 4: Apply Strategy
-**Action**: Write selected strategy to `memory/MEMORY_STREAM.md`:
-```
-META_STRATEGY: [Selected Strategy]
-REASON: [Justification]
-```
-
-### Step 5: Output
-**Action**: Display to user: "🚀 Strategy Selected: **[Strategy]** (based on [Reason])"
-
----
-
-## Exit Criteria
-**This phase is complete when ALL of the following are true**:
-- ✅ Strategy selected and logged.
-
----
-
-## Related Phases
-**Next Phase**: Phase 1
+# Phase 0.5: Meta-Cognitive Strategy Selection
+> **Parent Workflow**: 02-SPARC-Specification
+> **Phase Number**: 0.5
+> **Objective**: Adapt workflow behavior based on past performance.
+
+---
+
+## Prerequisites
+**MUST be completed before this phase**:
+- ✅ Phase 0: Ancestral Audit
+
+---
+
+## Instructions
+
+### Step 1: Read Performance Metrics
+**Action**: Read `memory/PERFORMANCE_LOG.md`.
+
+### Step 2: Pattern Analysis
+**Action**: Calculate the following from the last 3 entries:
+1.  **Avg Loops**: (Sum of loops) / 3
+2.  **Avg Mirror Score**: (Sum of scores) / 3
+
+### Step 3: Strategy Selection
+**Action**: Select one strategy strategy based on the data:
+
+| Condition | Strategy | Behavior |
+|-----------|----------|----------|
+| Avg loops > 5 | **VERBOSE** | Extra logging, more checkpoints |
+| Mirror Score < 8 | **CAUTIOUS** | Self-critique after every action |
+| BMO/Test Failures > 1 | **PARANOID** | Double-verify all assertions |
+| All metrics good | **STREAMLINED** | Skip optional checkpoints |
+| No Data | **STANDARD** | Default behavior |
+
+### Step 4: Apply Strategy
+**Action**: Write selected strategy to `memory/MEMORY_STREAM.md`:
+```
+META_STRATEGY: [Selected Strategy]
+REASON: [Justification]
+```
+
+### Step 5: Output
+**Action**: Display to user: "🚀 Strategy Selected: **[Strategy]** (based on [Reason])"
+
+---
+
+## Exit Criteria
+**This phase is complete when ALL of the following are true**:
+- ✅ Strategy selected and logged.
+
+---
+
+## Related Phases
+**Next Phase**: Phase 1