Documentation
CLI Reference

CLI Reference

Complete reference for all Conduct v0.1 CLI commands. Conduct provides 11 commands for complete memory management and orchestration.

Core Commands

conduct init

Initialize Conduct in your project with database setup and agent templates.

Usage:

conduct init

Interactive Prompts:

  1. Database Mode: Choose local (embedded libSQL) or remote (Turso)
  2. Profile Configuration: Set up database credentials
  3. Directory Creation: Initialize _conduct/ structure
  4. Agent Templates: Deploy 9 agent command templates

What It Creates:

  • _conduct/ directory structure with subdirectories
  • conduct.json configuration file
  • conduct.track.json local tracker
  • ~/.conduct/credentials global profile
  • Agent templates in .cursor/, .claude/, .warp/, .factory/
  • memory.db local database (if local mode)

Example:

$ conduct init
? Database mode: Local (embedded libSQL)
✅ Created _conduct/ directory
✅ Initialized memory.db
✅ Deployed 9 agent templates
✅ Configuration saved to conduct.json

conduct save

Execute SQL operations from file with validation.

Usage:

conduct save <file> [options]

Options:

  • -y, --yes - Skip confirmation prompt
  • --dry-run - Validate without executing

Features:

  • Validates SQL syntax and security
  • Whitelists INSERT/UPDATE only
  • Blocks DELETE, DROP, ALTER
  • Prevents SQL injection
  • Shows statement count before execution

Example:

$ conduct save _conduct/operations/update-run-1.sql

📝 SQL File: _conduct/operations/update-run-1.sql
✓ Validated 4 statements
  - 1 INSERT into run
  - 2 INSERT into run_feature
  - 1 UPDATE to run

? Execute SQL? (y/N) y
✅ Executed successfully
✅ 4 statements completed

conduct list

Query memory database and display all entities.

Usage:

conduct list [options]

Options:

  • -s, --specs - Show only specs
  • -r, --runs - Show only runs
  • -c, --checks - Show only checks
  • -f, --features - Show only features
  • -p, --packages - Show only packages
  • -a, --all - Show everything (default)
  • --json - Output as JSON
  • --status <status> - Filter by status
  • --limit <n> - Limit results (default: 20)

Example:

$ conduct list --specs --limit 5

📋 Specs (3):
  spec-1 [completed] Authentication System (simple)
    Location: _conduct/specs/1.v0.spec.md
    Agent: claude | LOE: simple
    
  spec-2 [in-progress] Dashboard UI (medium)
    Location: _conduct/specs/2/
    Agent: cursor | LOE: medium
    
  spec-3 [pending] API Integration (complex)
    Location: _conduct/specs/3.v0.spec.md
    Source: github#123

conduct health

Check system health and configuration.

Usage:

conduct health

Checks:

  • Database connection status
  • Configuration file validity
  • Schema version
  • Memory statistics
  • Credential profile

Example:

$ conduct health

🏥 Conduct Health Check

✅ Database: Connected
   Profile: default
   URL: file://_conduct/memory.db
   
✅ Configuration: Valid
   File: conduct.json
   Version: 0.1.0
   
✅ Schema: Up to date
   Version: 1.0.0
   
📊 Memory Statistics:
   Features: 15
   Specs: 8
   Runs: 12
   Checks: 10

conduct upgrade

Upgrade Conduct instructions and templates to latest version.

Usage:

conduct upgrade

What It Updates:

  • _conduct/AGENTS.md master instruction file
  • Root-level agent instructions (CLAUDE.md, WARP.md, GEMINI.md)
  • All 9 agent templates in _conduct/templates/
  • Agent-specific command files in .cursor/, .claude/, .warp/, .factory/, .opencode/
  • Preserves custom content outside managed blocks

Example:

$ conduct upgrade

🔄 Upgrading Conduct...

✅ Updated _conduct/AGENTS.md
✅ Updated CLAUDE.md
✅ Updated WARP.md
✅ Updated GEMINI.md
✅ Updated 36 template(s) in agent directories

✨ Upgrade complete! All instructions and templates are up to date.

Integration Commands

conduct sync

Sync with external issue trackers (GitHub, Linear).

Usage:

conduct sync [options]

Options:

  • --github-token <token> - GitHub API token
  • --linear-token <token> - Linear API token

Features:

  • Fetches issues from GitHub/Linear
  • Updates issue_connection table
  • Syncs spec/run status
  • Links entities to issues

Example:

$ conduct sync --github-token ghp_xxxxx

🔄 Syncing with GitHub...
✅ Fetched 15 issues
✅ Updated 8 connections
✅ Synced 3 spec statuses

📊 Summary:
   New connections: 5
   Updated connections: 3
   Failed: 0

conduct config

Manage configuration, database profiles, and issue tracker integration.

Usage:

conduct config [action] [subaction]

Commands:

# Profile management
conduct config profile list           # List all profiles
conduct config profile add <name>     # Add new profile
conduct config profile switch <name>  # Switch active profile

# Database operations
conduct config db test                # Test connection
conduct config db info                # Show connection details

# Issue tracker integration
conduct config tracker add            # Add GitHub or Linear
conduct config tracker remove         # Remove tracker config
conduct config tracker enable         # Enable a tracker
conduct config tracker disable        # Disable a tracker
conduct config tracker list           # Show configured trackers

Example - Profile Management:

$ conduct config profile list

📋 Database Profiles:

* default (active)
  URL: file://_conduct/memory.db
  Type: local
  
  production
  URL: libsql://conduct.turso.io
  Type: remote

$ conduct config profile switch production
✅ Switched to profile: production

Example - Issue Tracker Integration:

$ conduct config tracker add

? Select tracker: GitHub
? Enable immediately? Yes
? Enable automatic sync? No
? Default repository: 21nOrg/conduct

✅ GitHub tracker configured

Set GitHub token with:
  export GITHUB_TOKEN="your-token"

$ conduct config tracker list

⚙️  Issue Tracker Configuration

GitHub:
  Status: enabled
  Auto-sync: no
  Default repo: 21nOrg/conduct

Linear:
  Status: disabled

Discovery & Maintenance Commands

conduct discover

Automatically discover features in codebase.

Usage:

conduct discover [options]

Options:

  • --root-dir <dir> - Root directory to scan (default: current)
  • --package <name> - Package name to assign features to
  • --min-confidence <n> - Minimum confidence threshold 0-1 (default: 0.5)
  • -y, --yes - Auto-approve all discovered features
  • --json - Output as JSON

Strategies:

  1. Directory structure analysis (confidence: 0.8)
  2. Export analysis for TypeScript/JavaScript (confidence: 0.7)
  3. package.json hints (confidence: 0.9)

Example:

$ conduct discover --min-confidence 0.7

🔍 Discovering features...

📦 Discovered Features (8):

  authentication [90%]
    Paths: src/auth, lib/auth
    Strategy: directory + exports
    
  user-profile [85%]
    Paths: src/profile
    Strategy: directory
    
  dashboard [75%]
    Paths: src/dashboard
    Strategy: directory

? Save 8 features to memory? (y/N) y

✅ Generated SQL: _conduct/operations/discovered-features.sql
💡 Run: conduct save _conduct/operations/discovered-features.sql

conduct reconcile

Detect drift between memory and codebase reality.

Usage:

conduct reconcile [options]

Options:

  • --root-dir <dir> - Root directory to scan (default: current)
  • --since-days <n> - Check changes from N days ago (default: 7)
  • --since-commit <hash> - Check changes since specific commit
  • -y, --yes - Auto-approve all suggested updates
  • --dry-run - Show suggestions without applying
  • --json - Output as JSON

Detection Methods:

  • File existence verification
  • Git log analysis
  • Git rename detection
  • Heuristic analysis

Example:

$ conduct reconcile --since-days 30

🔄 Analyzing changes since 30 days ago...

✓ Loaded 45 features
✓ Analyzed 127 commits
✓ Detected 3 changes

🔄 Suggested Updates:

  old-auth (active) → removed [95%]
    All feature paths deleted
    
  user-profile (active) → renamed [90%]
    Feature renamed: src/profile -> src/user-profile
    
  legacy-api (active) → deprecated [60%]
    Significant code removal (350 lines deleted)

? Apply 3 suggested updates? (y/N) y

✅ Generated SQL: _conduct/operations/reconciled-updates.sql
💡 Run: conduct save _conduct/operations/reconciled-updates.sql

conduct relevancy

Calculate and manage relevancy scores for runs.

Usage:

conduct relevancy [options]

Options:

  • --run-id <id> - Calculate for specific run
  • --show-archivable - Show runs that should be archived
  • --recalculate-all - Recalculate scores for all runs
  • --json - Output as JSON

Formula:

score = baseScore * e^(-days/halfLife) * (1 + boost * references)

Parameters (configurable in conduct.json):

  • halfLife: 180 days (default)
  • referenceBoost: 0.1 (10% per reference)
  • archiveThreshold: 0.1 (10%)

Example:

$ conduct relevancy --show-archivable

📊 Run Relevancy Scores:

  run-1 [98%] active
    2 days old, 0 references
    
  run-2 [145%] active
    30 days old, 5 references (50% boost)
    
  run-3 [8%] 📦 archivable
    730 days old, 0 references
    
  run-4 [5%] 📦 archivable
    1095 days old, 0 references

💡 2 runs below archive threshold (10%)
💡 Run: conduct archive -y

conduct archive

Archive old runs based on relevancy score.

Usage:

conduct archive [options]

Options:

  • --threshold <n> - Archive threshold 0-1 (default: 0.1)
  • -y, --yes - Auto-approve archiving
  • --dry-run - Show what would be archived
  • --json - Output as JSON

Example:

$ conduct archive --dry-run

📦 Archivable Runs (score < 10%):

  run-3 - spec-123 [8%] 
    730 days old, 0 references
    Location: _conduct/runs/3.v0.run.md
    
  run-4 - spec-456 [5%]
    1095 days old, 0 references
    Location: _conduct/runs/4/

? Archive 2 runs? (y/N) y

✅ Archived 2 runs
✅ Updated relevancy scores
✅ Run IDs archived: 3, 4

Configuration Files

conduct.json

Project configuration file with JSON schema support:

{
  "$schema": "https://docs.conduct.run/schema/conduct.json",
  "version": "0.1.0",
  "memory": {
    "profile": "default",
    "relevancy": {
      "enabled": true,
      "halfLife": 180,
      "referenceBoost": 0.1,
      "maxBoost": 1.0,
      "archiveThreshold": 0.1
    }
  },
  "trackers": {
    "github": {
      "enabled": true,
      "autoSync": false,
      "defaultRepo": "21nOrg/conduct"
    },
    "linear": {
      "enabled": false,
      "autoSync": false,
      "workspace": "acme-corp"
    }
  }
}

Schema Features:

  • IDE autocomplete and validation
  • Real-time error detection
  • Property documentation on hover
  • Available at https://docs.conduct.run/schema/conduct.json

~/.conduct/credentials

Global credentials file (INI format):

[default]
url = file://_conduct/memory.db
token = 

[production]
url = libsql://conduct.turso.io
token = eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9...

[staging]
url = libsql://conduct-staging.turso.io
token = eyJhbGciOiJFZERTQSIsInR5cCI6IkpXVCJ9...

conduct.track.json

Fast local tracker for quick lookups with JSON schema support:

{
  "$schema": "https://docs.conduct.run/schema/track.json",
  "specs": [
    {
      "id": "1",
      "location": "_conduct/specs/1.spec.md",
      "status": "completed",
      "version": 0
    }
  ],
  "runs": [
    {
      "id": "1",
      "spec_id": "1",
      "location": "_conduct/runs/1.run.md",
      "status": "completed"
    }
  ],
  "checks": [],
  "last_updated": "2025-11-03T10:00:00Z"
}

Schema Features:

  • IDE autocomplete for all fields
  • Validation for status enums
  • ISO 8601 date-time validation
  • Available at https://docs.conduct.run/schema/track.json

Agent Templates

Conduct provides 9 agent command templates deployed to _conduct/templates/:

  1. conduct-spec.md - Create specification
  2. conduct-design.md - Create UI design
  3. conduct-run.md - Execute specification
  4. conduct-check.md - Verify implementation
  5. conduct.md - Lightweight mode (no ceremony)
  6. conduct-index.md - Discover features
  7. conduct-reconcile.md - Sync memory with codebase
  8. conduct-dry-run.md - Plan without implementing
  9. conduct-dry-check.md - Verify plan without code check

These templates are copied to agent-specific directories (.cursor/commands/, .claude/commands/, etc.) during initialization.

Memory Update Workflow

Security by Separation

Conduct uses a two-step process for memory updates:

  1. Agent generates SQL (untrusted)
  2. Human validates and executes (trusted)

Example Workflow 

# Step 1: Agent generates SQL during run
_conduct/operations/update-run-1.sql

# Step 2: Human reviews
cat _conduct/operations/update-run-1.sql

# Step 3: Human validates and executes
conduct save _conduct/operations/update-run-1.sql

# Step 4: Verify
conduct list --runs

SQL Validation

The validator whitelists:

  • ✅ INSERT statements
  • ✅ UPDATE statements
  • ✅ Specific tables only
  • ❌ DELETE, DROP, ALTER blocked
  • ❌ Subqueries blocked
  • ❌ Function calls blocked
  • ❌ SQL injection patterns blocked

Best Practices

Daily Workflow 

# Morning: Check system health
conduct health

# During work: Use agent templates
# - conduct-spec for new features
# - conduct-run for implementation
# - conduct-check for verification

# After each run: Save to memory
conduct save _conduct/operations/update-run-X.sql

# End of day: Review memory
conduct list

Weekly Maintenance 

# Update relevancy scores
conduct relevancy --recalculate-all

# Archive old runs
conduct archive -y

# Check for drift
conduct reconcile

# Verify health
conduct health

Monthly Maintenance 

# Full reconciliation
conduct reconcile --since-days 30 -y

# Discover new features
conduct discover -y

# Aggressive archiving
conduct archive --threshold 0.15 -y

# Sync with issue trackers
conduct sync --github-token $GITHUB_TOKEN
Getting StartedAI Agent Templates