shynlee04/hivemind-plugin/hivemind-architect-strategist

// Deep hierarchy scan
scan_hierarchy({ action: "analyze" })
// Recall past decisions
recall_mems({ query: "architecture" })
// Save strategic insights
save_mem({
  shelf: "strategy",
  content: "Key insight...",
  tags: "architecture,planning"
})
// Immutable constraints
save_anchor({
  key: "arch_constraint_1",
  value: "Tools must be ≤100 lines"
})

// Pattern search
grep({ pattern: "import.*from", include: "*.ts" })
// File discovery
glob({ pattern: "src/**/*.ts" })
// Content reading
read({ filePath: "src/lib/example.ts" })
// Web research
websearch({ query: "best practices for X" })

// Analyze hierarchy and context state
scan_hierarchy({ action: "analyze" })
// Pattern detection across codebase
grep({ pattern: "import.*from", include: "*.ts" })
grep({ pattern: "TODO|FIXME|HACK", include: "*.ts" })
// Structure analysis
glob({ pattern: "**/*.ts" })
glob({ pattern: "src/tools/**/*.ts" })
glob({ pattern: "src/lib/**/*.ts" })

# Verify layer boundaries
find src/tools -name "*.ts" -exec wc -l {} \; | awk '$1 > 100 {print "VIOLATION: " $2 " has " $1 " lines"}'
 
# Check for LLM prompts in lib/
grep -r "prompt\|generateText\|streamText" src/lib/ && echo "VIOLATION: LLM calls in pure lib"
 
# Check for business logic in tools
grep -r "async.*=>" src/tools/*.ts | grep -v "schema\|lib\." && echo "WARNING: Business logic in tool"

Layer           | Constraint                      | Detection
----------------|---------------------------------|--------------------------------
tools/          | ≤100 lines, Zod schema + lib   | wc -l > 100, missing schema
lib/            | Pure TS, no LLM prompts         | grep for ai-sdk imports
hooks/          | Read-only, inject context       | grep for write operations
schemas/        | Zod validation + FK constraints| missing .refine() or .transform()

// Find circular dependencies
grep -r "from '\.\." src/ | sort | uniq -c | sort -rn
// Identify tight coupling
grep -r "import.*from.*src/" src/ | cut -d: -f1 | sort | uniq -c | sort -rn

# Quick debt scan
grep -rn "TODO\|FIXME\|HACK\|XXX" src/ | wc -l
 
# Debt severity assessment
grep -rn "any\|unknown\|@ts-ignore" src/ | wc -l
 
# Test coverage gaps
find tests -name "*.test.ts" | wc -l vs find src -name "*.ts" | wc -l

┌─────────────────────────────────────────────────────────────────┐
│                     SYNTHESIS PIPELINE                          │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  1. SCAN     →  Gather raw data                                 │
│     • scan_hierarchy for context state                          │
│     • glob/grep for code patterns                               │
│     • read files for implementation details                     │
│                                                                 │
│  2. FILTER   →  Remove noise, focus on signal                   │
│     • Exclude node_modules, dist, generated files              │
│     • Prioritize src/, tests/, docs/                            │
│     • Focus on recent changes (git log)                         │
│                                                                 │
│  3. ANALYZE  →  Find patterns, gaps, issues                     │
│     • Compare against architecture taxonomy                     │
│     • Check dependency graphs for cycles                        │
│     • Identify code smells and debt                             │
│                                                                 │
│  4. SYNTHESIZE → Create actionable insights                     │
│     • Group related issues                                      │
│     • Prioritize by impact/effort                               │
│     • Generate recommendations with evidence                    │
│                                                                 │
│  5. REPORT   →  Evidence-based recommendations                  │
│     • Cite specific files and lines                             │
│     • Provide verification commands                             │
│     • Include rollback plans                                    │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

✓ ≤100 lines per file
✓ Zod schema for input validation
✓ Delegates business logic to lib/
✓ Returns structured result object
✗ No direct database access
✗ No LLM prompt construction
✗ No complex business logic

✓ Pure TypeScript functions
✓ No LLM SDK imports
✓ No external API calls (except via abstraction)
✓ Fully testable without mocks
✗ No Zod parsing (belongs in tools)
✗ No file writes (belongs in tools via persistence lib)

✓ Read-only operations
✓ Inject context into conversation
✓ Return structured context object
✗ No write operations
✗ No side effects
✗ No LLM calls

✓ Zod validation with .refine()/.transform()
✓ FK constraints for related entities
✓ Clear error messages
✗ No business logic
✗ No file I/O

## Architecture Review: [Component/Module]
 
### Executive Summary
[1-2 sentences on overall health]
 
### Evidence Gathered
-Files scanned: [count]
-Patterns detected: [list]
-Debt items found: [count by severity]
 
### Findings
 
#### Critical (Must Fix)
1.[Issue] - [Evidence: file:line] - [Impact]
 
#### High Priority
1.[Issue] - [Evidence] - [Impact]
 
#### Medium Priority  
1.[Issue] - [Evidence] - [Recommendation]
 
### Recommendations
1.[Action] - [Why] - [Verification command]
 
### Rollback Plan
[If recommendations fail, how to revert]

# Architecture health check
find src -name "*.ts" -exec wc -l {} \; | awk '$1 > 100 {print}'
 
# Dependency analysis
grep -r "from '\.\." src/ | sort | uniq -c | sort -rn | head -20
 
# Technical debt scan
grep -rn "TODO\|FIXME\|any\|@ts-ignore" src/ | wc -l
 
# Test coverage check
npm test -- --coverage 2>/dev/null | grep "All files"
 
# Type safety
npx tsc --noEmit