npm i codeimpact

# Install globally
npm i -g codeimpact

# Or use with npx (no install)
npx codeimpact init

# Initialize in your project
cd your-project
codeimpact init

Outcome recorded → Diagnose (15-rule decision table)
  → wrong-assumption + fix → Patch SKILL.md immediately
  → wrong-assumption + fix + no SKILL.md → Patch matching knowledge skill's Watch Out
  → outdated-research → Flag for re-research
  → 2+ similar failures → Write lesson to AGENT.md (quarantine)
  → 3+ confirmations → Promote lesson to SKILL.md Rules
  → 90 days unconfirmed → Decay lesson

your-project/
├── .code-impact/                  # Agent workspace (auto-generated)
│   ├── config.yaml                # Configuration
│   ├── index.json                 # Manifest with tech, features, outcomes
│   ├── project/                   # Project-wide knowledge
│   │   ├── SKILL.md               # Technology stack + conventions
│   │   ├── CONVENTIONS.md         # Coding standards
│   │   ├── ARCHITECTURE.md        # System architecture
│   │   └── AGENT.md               # Super agent (coordinator)
│   ├── research/                  # Distilled tech documentation
│   │   └── express@4.21.0.md      # With trust scores + structured sections
│   └── features/                  # Feature-specific knowledge
│       └── auth/
│           ├── SKILL.md           # Feature rules, pitfalls, research refs
│           └── AGENT.md           # Scope, tools, lessons learned
├── AGENTS.md                      # Top-level agent directory (auto-generated)
├── knowledge/                     # Legacy knowledge workspace
│   └── skills/                    # AI-created SKILL.md files
├── .codeimpact/
│   └── codeimpact.db              # SQLite database
└── src/

codeimpact agents init             # Initialize agent workspace
codeimpact agents generate         # Run full pipeline (detect → research → generate)
codeimpact agents status           # Show agent system health + lifecycle proposals
codeimpact agents research         # Refresh technology research
codeimpact agents migrate          # Migrate knowledge/ → .code-impact/

---
name: better-sqlite3-patterns
description: Synchronous database patterns. Use when working with database queries.
version: 1.0
metadata:
  scope: technology
  created_by: ai
---

# better-sqlite3 Patterns

## When to Use
When modifying database queries, adding tables, or working with any file
that imports from src/storage/database.ts.

## Rules
- ALL database access goes through database.ts — never import better-sqlite3 directly.
- Use db.prepare().all() for SELECT, db.prepare().run() for INSERT/UPDATE/DELETE.

## Pitfalls
- db.exec() returns nothing. If you use it for SELECT, you get undefined.
- better-sqlite3 is synchronous. Do NOT wrap in async/await.

## Verification
- npx tsc --noEmit passes with no type errors on database code.

codeimpact deadcode
codeimpact deadcode --json --threshold 80

Dead Code Report:
- 4,230 lines of unused code detected
- 12 files with zero imports
- 23 functions never called
Safe to delete: 89% confidence

codeimpact test-impact
codeimpact test-impact --changed src/auth/login.ts
codeimpact test-impact --branch main

Analyzing impact of src/auth/login.ts...
Files affected: 12
Tests to run: 8 (instead of 234)
Estimated time: 2m (instead of 28m)
Time saved: 26 minutes

codeimpact impact src/core/engine.ts
codeimpact impact src/auth/session.ts --depth 5 --json

File: src/auth/session.ts
Risk Score: 78/100 (HIGH)
Direct dependents: 8 files
Transitive dependents: 34 files
Critical paths affected: src/api/checkout.ts, src/billing/payments.ts
Recommendation: Senior review required

codeimpact stats
codeimpact stats --period week
codeimpact stats --period all --json

This Month:
- Queries: 1,247
- Tokens used: 892K
- Estimated cost: $5.35

"What files depend on src/auth/login.ts?"
"Show me the import chain for this module"

"If I change this file, what else might break?"
"What's the blast radius of changing this module?"
"What tests cover this function?"

"Check memory_status for knowledge gaps"
"Create a skill for the authentication patterns I just implemented"
"What skills exist for the database layer?"

"List all agents and their scopes"
"Validate if auth-agent can modify src/billing/pay.ts"
"Show the telemetry dashboard for this week"
"What lifecycle proposals exist?"

"Are there any unused exports in this project?"
"Which files have no dependents?"

"Find all authentication-related code"
"Where is the user validation logic?"

"Are there any circular dependencies?"
"What decisions have we made about authentication?"

# Setup
codeimpact init              # Set up project + configure AI tools
codeimpact serve             # Start HTTP API server

# Analysis
codeimpact deadcode          # Find unused exports and dead code
codeimpact test-impact       # Find which tests to run for changes
codeimpact impact <file>     # Analyze blast radius of a file
codeimpact stats             # Show token usage and costs
codeimpact reindex           # Force reindex after git operations

# Agent System
codeimpact agents init       # Initialize agent workspace
codeimpact agents generate   # Detect → research → generate → learn
codeimpact agents status     # Health, outcomes, lifecycle proposals
codeimpact agents research   # Refresh technology research
codeimpact agents migrate    # Migrate knowledge/ → .code-impact/

# Project Management
codeimpact projects list     # List registered projects
codeimpact projects add .    # Add current directory
codeimpact projects switch   # Switch active project
codeimpact export            # Export decisions to ADR files
codeimpact help              # Show help

--project, -p <path>      # Path to the project directory
--json                    # Output as JSON
--threshold <percent>     # Minimum confidence % (for deadcode)
--changed <file>          # Specify changed file(s) (for test-impact)
--git-diff                # Use git diff to detect changes
--branch <name>           # Compare to branch (e.g., main)
--depth <n>               # Max dependency depth (default: 3)
--period <type>           # Time period: day, week, month, all

codeimpact serve --project /path/to/project --port 3333

curl "http://localhost:3333/impact?file=src/auth/login.ts"

{
  "mcpServers": {
    "codeimpact": {
      "command": "npx",
      "args": ["-y", "codeimpact", "--project", "/path/to/your/project"]
    }
  }
}

your-project/
├── .codeimpact/
│   ├── codeimpact.db        # SQLite database (index, outcomes, telemetry)
│   ├── tier1.json           # Hot context cache
│   └── feature-context.json # Session tracking
├── .code-impact/            # Agent workspace (auto-generated)
│   ├── config.yaml          # Agent system configuration
│   ├── index.json           # Technologies, features, outcomes, lifecycle
│   ├── project/             # Project-wide SKILL/CONVENTIONS/ARCHITECTURE/AGENT
│   ├── research/            # Distilled tech docs with trust scores
│   └── features/            # Per-feature SKILL + AGENT files
├── knowledge/               # Knowledge workspace
│   ├── skills/              # SKILL.md files (auto-generated + AI-created)
│   │   ├── technology/      # Tech pattern skills (auto-generated)
│   │   ├── features/        # Feature-scoped skills (auto-generated)
│   │   └── core/            # Project conventions (auto-generated)
│   ├── docs/                # Generated documentation
│   └── index.json           # Knowledge manifest (includes autoGeneration tracking)
├── src/
└── ...

~/.codeimpact/
└── registry.json           # Project list

git clone https://github.com/abhisavakar/codeimpact.git
cd codeimpact
npm install
npm run build
npm test              # 180+ tests including 10 eval scenarios

src/core/agents/                  # Super Agent System (12 modules)
├── orchestrator.ts               # Pipeline: init → detect → research → generate → learn
├── tech-detector.ts              # Lockfile + manifest scanning
├── feature-detector.ts           # Cohesion clustering + CODEOWNERS
├── research-engine.ts            # Fetch + distill + trust scoring
├── generator.ts                  # SKILL.md / CONVENTIONS.md / ARCHITECTURE.md
├── agent-generator.ts            # AGENT.md + AGENTS.md shim
├── improvement-engine.ts         # Self-learning: diagnose → learn → promote → decay + knowledge skill patching
├── diagnosis-table.ts            # 15-rule declarative error classification
├── outcome-storage.ts            # SQLite agent_outcomes table
├── lifecycle.ts                  # Merge/split/prune proposals
├── telemetry.ts                  # Event emission + dashboard queries
├── marker-writer.ts              # Content-hash dedup + token budgets
├── token-budget.ts               # Section-priority truncation
├── migration.ts                  # knowledge/ → .code-impact/ migration
├── research-trust.ts             # 7-signal trust scoring
├── remote-provider.ts            # GitHub/GitLab/Bitbucket abstraction
├── co-change-analyzer.ts         # Git log analysis for feature coupling
└── workspace.ts                  # Directory management + config

src/core/knowledge/               # Knowledge Layer
├── orchestrator.ts               # Knowledge generation pipeline (intel → providers → features → auto-skills → evolution)
├── skill-auto-generator.ts       # Auto-generates starter skills from codebase data (NEW in 0.5.0)
├── skill-generator.ts            # SKILL.md rendering + writing
├── skill-reader.ts               # Read/search skills across all scopes
├── skill-evolution.ts            # Evolve existing skills from usage data
├── intelligence-collector.ts     # Collect project intelligence (techs, risks, patterns)
├── provider-research.ts          # Technology documentation research
├── platform-sync.ts              # Sync rules to IDE platform files
├── doc-sync.ts                   # Sync architecture/component/changelog docs
└── workspace.ts                  # Knowledge workspace paths + manifest I/O