hive-mind-advanced

---

LIBRARY-FIRST PROTOCOL (MANDATORY)

Before writing ANY code, you MUST check:

Step 1: Library Catalog

• Location: .claude/library/catalog.json
• If match >70%: REUSE or ADAPT

Step 2: Patterns Guide

• Location: .claude/docs/inventories/LIBRARY-PATTERNS-GUIDE.md
• If pattern exists: FOLLOW documented approach

Step 3: Existing Projects

• Location: D:\Projects\*
• If found: EXTRACT and adapt

Decision Matrix

Match	Action
Library >90%	REUSE directly
Library 70-90%	ADAPT minimally
Pattern exists	FOLLOW pattern
In project	EXTRACT
No match	BUILD (add to library after)

---

STANDARD OPERATING PROCEDURE

Purpose

Operate advanced hive-mind swarms that blend exploration and exploitation, maintain shared memory, and deliver consensus-backed outputs without overclaiming.

Trigger Conditions

• Positive: collective reasoning, divergent/convergent rounds, consensus/selection, shared-memory swarms, conflict resolution among agents.
• Negative: isolated single-agent tasks, fixed pipelines without deliberation, or prompt-only edits (route to prompt-architect).

Guardrails

• Skill-Forge structure-first: maintain SKILL.md, examples/, tests/; provision resources/ and references/ or document gaps.
• Prompt-Architect hygiene: collect intent, constraints, and perspectives; tag HARD/SOFT/INFERRED; output pure English with ceiling-noted confidence.
• Collective safety: define roles (scouts, synthesizers, judges), prevent groupthink via adversarial reviewer, and enforce registry use plus hook latency budgets.
• Adversarial validation: run dissent and tie-break drills, conflict resolution tests, and COV loops; capture evidence.
• MCP tagging: persist hive runs with WHO=hive-mind-advanced-{session} and WHY=skill-execution.

Execution Playbook

1. Intent & framing: clarify objective, success metrics, and diversity requirements; confirm inferred constraints.
2. Swarm design: assign roles, cadence (diverge/converge), and memory layout; set evidence requirements.
3. Deliberation rounds: collect proposals, adversarial reviews, and synthesis; manage quotas and timeboxes.
4. Decision & commit: select outputs via criteria, record dissent, and define rollback/next-iteration triggers.
5. Validation loop: stress-test consensus, latency, and bias; run COV and store telemetry.
6. Delivery: summarize chosen path, dissenting views, evidence, residual risks, and confidence ceiling.

Output Format

• Objective, constraints, and swarm roles.
• Deliberation cadence, decision criteria, and memory approach.
• Evidence log (supporting and dissenting), risks, and follow-ups.
• Confidence: X.XX (ceiling: TYPE Y.YY) - rationale.

Validation Checklist

• Structure-first assets present or ticketed; examples/tests reflect hive patterns.
• Diversity and dissent enforced; registry and hooks verified; rollback/iteration rules documented.
• Adversarial/COV runs stored with MCP tags; confidence ceiling stated; English-only output.

Completion Definition

Swarm decision is complete when consensus is reached (or documented disagreement), evidence is stored, risks are owned, and next actions or rollbacks are clear.

Confidence: 0.70 (ceiling: inference 0.70) - Hive-mind SOP now mirrors skill-forge rigor and prompt-architect evidence/confidence discipline.