Commit graph

6 commits

Author SHA1 Message Date
Danny Stocker
8544f5a020 Session 5: Active Quality Assurance Partner (assist Sessions 1-4)
**NEW AGENTS (Start Immediately - NO Dependencies):**

Agent 0A: Evidence Quality Standards Deployment (CRITICAL - First 10min)
- Deploys EVIDENCE_QUALITY_STANDARDS.md for Sessions 1-4
- Citation format templates (IF.TTT compliance)
- Evidence quality scoring rubric (primary/secondary/tertiary sources)
- Multi-source verification examples
- Confidence score guidelines (0.95+ requires ≥2 primary sources)

Agent 0B: Real-Time Quality Monitor (CONTINUOUS - Every 5min)
- Polls intelligence/session-*/ for new commits
- Reviews citations for IF.TTT compliance (SHA-256, ≥2 sources, line numbers)
- Creates QUALITY_FEEDBACK.md (updated every 5min)
- Sessions 1-4 read feedback → fix issues proactively (prevent rework)
- ESCALATE if >20% citations lack compliance

Agent 0C: Guardian Briefing Templates (PREP WORK)
- Creates 20 guardian-specific briefing templates
- Consensus prediction formula (evidence quality 40%, multi-source 30%, feasibility 20%, philosophy alignment 10%)
- Voting criteria checklists

**Benefits:**
- Zero idle time: Session 5 productive for full 3-hour window (not just 20min prep + 2h40min waiting)
- Prevent rework: Sessions 1-4 follow quality standards from start
- Faster validation: Session 5 familiar with evidence as it arrives (real-time review)
- Budget efficiency: $25 used for active QA (prevents expensive rework at validation stage)

**Phase 2 (Agents 1-10):** Evidence extraction & Guardian validation (wait for Sessions 1+2+3+4)

**InfraFabric S² Pattern:** Continuous feedback loop (3,563× faster than batch validation)

Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-13 03:03:11 +01:00
Danny Stocker
cb7ad641cd Fix absolute paths for cloud sessions (no shared filesystem)
**Path Fixes for Multi-Machine Deployment:**

Session 2:
- Changed `/home/setup/navidocs/ARCHITECTURE_INTEGRATION_ANALYSIS.md` → `ARCHITECTURE_INTEGRATION_ANALYSIS.md`
- Changed `/home/setup/navidocs/YACHT_ACCOUNTING_RESEARCH.md` → `YACHT_ACCOUNTING_RESEARCH.md`

Session 5:
- Changed `/home/setup/infrafabric/schemas/citation/v1.0.schema.json` → `schemas/citation/v1.0.schema.json`
- Copied citation schema from infrafabric repo into navidocs repo (no external dependency)
- Fixed example citation paths (absolute → relative)

**Why This Matters:**
Cloud sessions run on separate machines with NO shared filesystem. All paths must be:
1. Relative to git repo root (e.g., `./file.md` not `/home/setup/navidocs/file.md`)
2. Files must exist IN the navidocs repo (not external repos)

**Test Command:**
```bash
git clone https://github.com/dannystocker/navidocs.git
cd navidocs
git checkout navidocs-cloud-coordination
# All referenced files now accessible via relative paths
```

Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-13 02:52:07 +01:00
Danny Stocker
da1263d1b3 Add IF.bus intra-agent communication protocol to all 5 cloud sessions
- Added IFMessage schema with FIPA-ACL performatives
- Session-specific communication flows (distributed intelligence, peer review, adversarial testing, sequential handoffs, consensus building)
- Automatic conflict detection (>20% variance triggers ESCALATE)
- Multi-source verification (IF.TTT ≥2 sources requirement)
- Token cost tracking (IF.optimise integration)
- PARALLEL_LAUNCH_STRATEGY.md for simultaneous session deployment
- SWARM_COMMUNICATION_PROTOCOL.md comprehensive protocol docs

Based on InfraFabric S² multi-swarm coordination (3,563x faster than git polling)

🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-13 02:03:24 +01:00
Danny Stocker
a5ffcb5769 Add agent identity & check-in protocol to all 5 sessions
Each session now has:
- Agent self-identification system (S1-H01 through S5-H10)
- Check-in protocol (agents announce identity at start)
- Task reference system (agents find instructions by number)
- Dependency tracking (parallel vs sequential execution)

Note: Prestige yachts sell for €1.5M (not €480K) - needs price correction in Session 1

🤖 Generated with Claude Code
2025-11-13 01:27:48 +01:00
Danny Stocker
317c01e871 Update Sessions 4-5: align with sticky engagement model
Session 4: Implementation plan for inventory tracking, cameras, maintenance, contacts, search UX
Session 5: Guardian validation of daily engagement model (not passive doc vault)

All 5 sessions now pivot to recreational boats (Jeanneau Prestige 40-50ft) with sticky daily-use features.

🤖 Generated with Claude Code
2025-11-13 00:58:48 +01:00
Danny Stocker
760fedc138 Add InfraFabric S2 multi-swarm integration plan for yacht sales intelligence
Created 5 cloud session starter prompts applying Epic V4 + S2 methodology:
- Session 1: Market research (10 Haiku agents, $15 budget)
- Session 2: Technical architecture (10 Haiku agents, $20 budget)
- Session 3: UX/Sales enablement (10 Haiku agents, $15 budget)
- Session 4: Implementation planning (10 Haiku agents, $15 budget)
- Session 5: Evidence synthesis + Guardian Council vote (10 Haiku agents, $25 budget)

Total budget: $90 ($10 buffer) across 50 Haiku agents + 5 Sonnet coordinators

Integration plan leverages:
- IF.bus (event-driven architecture)
- IF.swarm (multi-agent orchestration)
- IF.optimise (49.3% validated token cost reduction)
- IF.guard (20-voice Guardian Council consensus)
- IF.TTT (medical-grade evidence with ≥2 sources per claim)

Objective: Prepare Riviera Plaisance yacht sales pitch + 4-week NaviDocs enhancement roadmap (warranty tracking, Home Assistant integration, sale workflow automation)

🤖 Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-13 00:43:27 +01:00