docs: Phase 7 Final Validation & Complexity Update

.agent/workflows/multi_agent_audit.md

@@ -12,6 +12,7 @@ Mandatory for any change touching: order submission, position sizing, FSM state
 
 1. Identify the implementation to audit (file path, diff, or code block).
 2. Define audit criteria:
+   - **Systemic Anti-Patterns (MANDATORY)**: Mandate the grouping of localized flaws under root-cause architectural patterns (e.g., TOCTOU, Bypassing Enqueue, Non-Atomic mutations).
    - **Safety**: ghost orders, naked positions, shutdown races
    - **Correctness**: FSM state coverage, edge cases
    - **DNA compliance**: no locks, correct pattern (Enqueue vs direct-write), ASCII gate
@@ -23,6 +24,8 @@ Mandatory for any change touching: order submission, position sizing, FSM state
 
 Invoke each auditor with IDENTICAL input. They must NOT see each other's output.
 
+**CRITICAL PROMPT INSTRUCTION:** All auditors MUST audit the architectural contracts, not just the execution paths. Do not simply list individual bugs. You must synthesize and group every localized bug under a systemic structural anti-pattern or V12 DNA violation.
+
 | Auditor   | Tool            | Focus                                |
 | --------- | --------------- | ------------------------------------ |
 | FORENSICS | Codex forensics | Logic trace, state sequence proof    |
@@ -36,8 +39,12 @@ RED TEAM AUDIT — [Auditor Name] — [Date]
 
 VERDICT: [PASS / FAIL / CONDITIONAL]
 
+Systemic Anti-Patterns:
+1. [Pattern Name]: [Description] (Affects Findings X, Y)
+2. ...
+
 Findings:
-1. [Finding + severity: CRITICAL / WARNING / INFO]
+1. [Finding + severity: CRITICAL / WARNING / INFO] (Maps to Pattern X)
 2. ...
 
 Recommendation: [APPROVE / REVISE / BLOCK]

.bob/commands/bob-tdd.md

@@ -0,0 +1,185 @@
+---
+description: Execute a single complexity extraction ticket using the full P2-P6 TDD Red-Green-Refactor pipeline.
+argument-hint: <traycer-ticket-content>
+---
+# MISSION: Bob TDD -- V12 Photon Kernel Complexity Extraction
+**Spec Ref**: docs/brain/bob_tdd_workflow.md
+**Protocol**: V12 Photon Kernel DNA (Lock-Free, ASCII-Only, TDD Red-Green-Refactor)
+
+---
+
+## STEP 1 -- P2 FORENSICS (Plan Mode)
+
+Before writing any code or plan, run forensic analysis:
+
+### 1a. jCodemunch Structural Scan
+- `get_file_outline` on the target file -- map every symbol, signature, complexity score
+- `get_blast_radius` on the target method -- identify all downstream callers
+- `find_references` on any shared state accessed in the method
+
+### 1b. Complexity Baseline
+Run: `python scripts/complexity_audit.py`
+Record the BEFORE CYC score for the target method.
+
+### 1c. Graphify Caller Map
+Run: `graphify update .`
+Read `graphify-out/GRAPH_REPORT.md` -- confirm caller impact scope.
+
+**Output**: Write `docs/brain/forensics_report_t[ID].md` with:
+- Target method name + current CYC score
+- Blast radius (callers list)
+- Proposed extraction boundary (helper name + signature draft)
+- DNA risks identified
+
+---
+
+## STEP 2 -- P3 ARCHITECT PLAN (Plan Mode)
+
+Produce a written implementation plan:
+
+```
+## Bob TDD Plan: [ticket ID] -- [method name]
+### Extraction Design
+| Helper Name | Signature | Lines Extracted | CYC Impact |
+|-------------|-----------|-----------------|------------|
+| ...         | ...       | ...             | ...        |
+
+### Caller Impact
+| Caller File | Caller Method | Change Required |
+|-------------|---------------|-----------------|
+| ...         | ...           | ...             |
+
+### TDD Contract Tests Required
+| Test # | Scenario        | Expected Result |
+|--------|-----------------|-----------------|
+| 1      | Happy path      | ...             |
+| 2      | Null/guard edge | ...             |
+| 3      | Caller invariant| ...             |
+```
+
+**Output**: Write `docs/brain/implementation_plan_t[ID].md`
+
+### !!! DIRECTOR APPROVAL GATE !!!
+**STOP HERE. Do NOT proceed to Step 3 until the Director explicitly confirms.**
+
+Output: "[BOB-TDD-GATE] Plan written to docs/brain/implementation_plan_t[ID].md. Awaiting Director approval."
+
+---
+
+## STEP 3 -- P4 ADJUDICATOR AUDIT (Internal)
+
+Perform adversarial self-audit of the plan against V12 DNA:
+
+Checklist:
+- [ ] Zero lock() usage in proposed code
+- [ ] No Thread.Sleep in proposed tests
+- [ ] Extraction is >= 15 LOC (extraction floor)
+- [ ] No logic drift -- pure structural extraction
+- [ ] ASCII-only in all string literals
+- [ ] deploy-sync.ps1 is included in post-edit sequence
+
+**Output**: Write `docs/brain/adjudicator_audit_t[ID].md`
+
+If any checklist item FAILS: return `CONDITIONAL PASS` with specific clarification.
+If all pass: return `PASS -- CLEARED FOR P5 EXECUTION`.
+
+---
+
+## STEP 4 -- P5 ENGINEER (Advanced/Code Mode) -- RED-GREEN
+
+### RED Phase: Write Failing Contract Tests FIRST
+
+Before touching src/, write the contract tests to `tests/[SubgraphName]IntegrationTests.cs`:
+
+Required scenarios:
+1. **Happy path**: normal input -> expected extracted-helper output
+2. **Null/guard edge**: boundary condition -> must not throw or corrupt state
+3. **Caller invariant**: call site behavior is identical before and after extraction
+
+The tests MUST fail at this point (RED). Do NOT proceed if they pass -- that means
+the test is not actually targeting the new helper.
+
+### GREEN Phase: Extract the Method
+
+Apply surgical extraction:
+- Use `v12_split.py` for any extraction exceeding 50 lines (manual copy-paste BANNED)
+- Touch ONLY the target method and its new helper
+- NEVER mutate whitespace, indentation, or adjacent unrelated code
+- After extraction, run the contract tests -- they must now PASS (GREEN)
+
+**Self-healing retry**: If GREEN fails, re-examine extraction boundary and retry up to 3 times.
+If 3 attempts fail: HALT. Report exact failure trace. Do NOT proceed.
+
+### Post-Edit Deployment (MANDATORY)
+```powershell
+# Re-establish hard links + ASCII gate
+powershell -File .\deploy-sync.ps1
+
+# Lock regression audit (must return ZERO matches)
+grep -r "lock(" src/
+
+# Unicode regression audit (must return ZERO matches)
+grep -Prn "[^\x00-\x7F]" src/
+```
+
+All three must PASS before proceeding to P6.
+
+---
+
+## STEP 5 -- P6 VERIFIER (Plan/Code Mode) -- REFACTOR
+
+Run full verification suite:
+
+```powershell
+# Full test suite
+dotnet test tests/
+
+# Complexity audit -- confirm CYC delta meets ticket target
+python scripts/complexity_audit.py
+
+# Final hard-link sync
+powershell -File .\deploy-sync.ps1
+```
+
+**Output**: Write `docs/brain/verification_report_t[ID].md` containing:
+- Test pass rate (e.g., `20/20 PASS`)
+- CYC before/after delta
+- Lock audit: CLEAN
+- Unicode audit: CLEAN
+- deploy-sync.ps1: PASS
+- BUILD_TAG (bump now)
+
+---
+
+## STEP 6 -- HANDOFF TO DIRECTOR
+
+Only after ALL Step 5 audits PASS, output:
+
+```
+[BOB-TDD-COMPLETE]
+Ticket: [ID]
+Method: [target method]
+CYC: [before] -> [after]
+Tests: [N]/[N] PASS
+BUILD_TAG: [new tag]
+Status: READY FOR F5 COMPILE
+
+Director Post-Ticket Checklist:
+[ ] Press F5 in NinjaTrader -- verify BUILD_TAG banner
+[ ] Confirm complexity_audit.py pass in verification report
+[ ] Confirm test pass rate in verification report
+[ ] Commit forensics + plan + verification reports
+[ ] Update BUILD_TAG_BASELINE in next ticket's header prompt
+```
+
+---
+
+## BANNED PATTERNS (immediate halt)
+
+- `lock(anything)` -- BANNED
+- `Monitor.Enter` / `Monitor.Exit` -- BANNED
+- `Thread.Sleep()` anywhere -- BANNED
+- Unicode / emoji / curly quotes in any string literal -- BANNED
+- Manual copy-paste for extractions > 50 lines -- BANNED (use v12_split.py)
+- Skipping RED phase (writing GREEN without a failing test first) -- BANNED
+- Proceeding past any GATE without explicit Director confirmation -- BANNED

.bob/commands/bug-bounty-consolidate.md

@@ -0,0 +1,112 @@
+---
+description: Consolidate 7 cluster bug reports into a validated, hallucination-filtered, ranked repair plan. Run after all 7 cluster-bug-hunt reports are complete.
+argument-hint: runner (optional -- runner suffix, e.g. bob, qwen, jules, gemini. Defaults to bob)
+---
+# MISSION: Bug Bounty Consolidation -- V12 Photon Kernel
+**Spec Ref**: docs/brain/bug_bounty_workflow.md
+**Input**: docs/brain/bug_report_s1_[runner].md through bug_report_s7_[runner].md (all 7 must exist)
+**Output**: docs/brain/cluster_bug_bounty_report_[runner].md
+**Mode**: Plan mode -- READ ONLY. No src/ edits.
+
+---
+
+## STEP 1 -- INGEST ALL 7 REPORTS
+
+Read the following files in order:
+- docs/brain/bug_report_s1.md (S1: SIMA Core)
+- docs/brain/bug_report_s2.md (S2: Execution Engine)
+- docs/brain/bug_report_s3.md (S3: UI & Photon IO)
+- docs/brain/bug_report_s4.md (S4: REAPER Defense)
+- docs/brain/bug_report_s5.md (S5: Kernel State)
+- docs/brain/bug_report_s6.md (S6: Signals & Entries)
+- docs/brain/bug_report_s7.md (S7: Kernel Infrastructure)
+
+If any file is missing: HALT and report to Director which clusters are incomplete.
+
+---
+
+## STEP 2 -- HALLUCINATION FILTER
+
+For EVERY reported bug across all 7 reports:
+1. Use jCodemunch `search_symbols` to verify the cited method exists
+2. Use `get_file_content` to verify the cited code pattern matches actual src/
+3. Use `find_references` to confirm the cited shared state is actually accessed
+
+Disposition for each bug:
+- VALIDATED: evidence confirmed in src/
+- FILTERED: cited method/pattern does not match src/ reality
+- UNCERTAIN: partially verifiable -- flag for Director review
+
+Track and report the filter rate per cluster.
+
+---
+
+## STEP 3 -- CROSS-CLUSTER DEDUPLICATION
+
+Identify bugs reported by multiple agents for the same root cause:
+- Match on: same file + same method + same root cause mechanism
+- Merge into single canonical entry
+- List all clusters that reported it
+- Elevate severity if blast radius spans 2+ clusters
+
+---
+
+## STEP 4 -- SEVERITY RANKING
+
+Final ranking of all validated bugs:
+- Critical: Data corruption, race conditions, use-after-free
+- High: FSM state leaks, ghost order windows, O(N^2) hot paths, semaphore leaks
+- Med: Missing null guards, incomplete resets, inefficient lookups
+- Low: Style violations, minor inefficiencies
+
+---
+
+## STEP 5 -- OUTPUT
+
+Write docs/brain/cluster_bug_bounty_report_[runner].md containing:
+
+```
+# V12 Cluster Bug Bounty Report
+Generated: [date]
+
+## Summary
+Total bugs found (raw): [N]
+Validated: [N] | Filtered (hallucinations): [N] | Uncertain (Director review): [N]
+Critical: [N] | High: [N] | Med: [N] | Low: [N]
+
+## Filter Rate by Cluster
+| Cluster | Found | Validated | Filtered |
+|---------|-------|-----------|----------|
+| S1 SIMA | N | N | N |
+...
+
+## Validated Bug List (ranked by severity)
+[full list in BUG-[S#]-[NNN] format]
+
+## Recommended Repair Sequence
+[cluster order based on Critical count and dependency graph]
+
+## /epic-tdd Ticket Blocks
+[copy-paste ready ticket for each validated bug]
+```
+
+---
+
+## STEP 6 -- HANDOFF
+
+Output:
+```
+[BUG-BOUNTY-CONSOLIDATION-COMPLETE]
+Total validated: [N]
+Filtered: [N]
+Uncertain (needs Director review): [N]
+Report: docs/brain/cluster_bug_bounty_report_[runner].md
+Next: Director reviews report -> selects cluster -> /epic-tdd for repairs
+```
+
+---
+
+## BANNED
+- Any src/ edit -- BANNED
+- Fixing bugs inline -- BANNED
+- Marking a bug VALIDATED without jCodemunch verification -- BANNED