Trinity Methodology Adoption
Convert any Claude Code agent to Trinity-compatible format following best practices.
code
ADOPT → DEVELOP → DEPLOY → RUN → MONITOR (THIS (Local) (Sync) (Remote) (Notify) SKILL)
Related Skills
| Skill | Purpose |
|---|---|
/trinity-compatibility | Read-only audit of current state |
/trinity-adopt | This skill - Implement changes to become Trinity-compatible |
/trinity-sync | Git-based synchronization with remote |
/trinity-remote | Remote agent operations (execute, deploy-run) |
/trinity-schedules | Scheduled task management |
Command Modes
Parse $ARGUMENTS to determine mode:
| Command | Mode | Description |
|---|---|---|
/trinity-adopt | adopt | Full adoption workflow (default) |
/trinity-adopt analyze | analyze | Analysis only, no changes |
What is Trinity Methodology?
A maturity model for business agent development:
code
EXPLORE → ADOPT → CODIFY → SCHEDULE → AUTONOMOUS
(Manual) (You (Build (Trinity (Monitored
are policies/ runs it) execution)
here) procedures)
Key principle: Skills, policies, and procedures ARE the agent. Everything else is runtime state.
Mode: Analyze Only (analyze)
Run analysis without making changes. Equivalent to /trinity-compatibility.
Workflow
- •Check current state
- •Generate compatibility report
- •List required changes (but don't implement)
Mode: Full Adoption (adopt - default)
Convert the current agent to Trinity-compatible format.
Step 1: Current State Analysis
bash
ls -la ls .claude/ 2>/dev/null ls .claude/skills/ 2>/dev/null ls -d .claude/skills/policy-* .claude/skills/procedure-* 2>/dev/null
Step 2: Present Adoption Report
code
## Trinity Adoption Analysis ### Current State - Agent directory: [path] - CLAUDE.md: [exists/missing] - .claude/skills/: [count] skills - Existing policies: [list or none] - Existing procedures: [list or none] ### Required for Adoption | Component | Status | Action Needed | |-----------|--------|---------------| | CLAUDE.md | [X]/[ ] | [action] | | template.yaml | [X]/[ ] | [action] | | .mcp.json.template | [X]/[ ] | [action] | | .env.example | [X]/[ ] | [action] | | .gitignore | [X]/[ ] | [action] | | outputs/ | [X]/[ ] | [action] | | scripts/ | [X]/[ ] | [action] |
Step 3: Confirm Adoption
Ask user: "Ready to adopt Trinity methodology? This will:
- •Create missing required files
- •Set up directory structure
- •Create initial policy: policy-agent-purpose
Proceed? [Y/n]"
Step 4: Execute Adoption
4a. Create Required Directories
bash
mkdir -p outputs scripts .claude/skills .claude/agents memory
4b. Create template.yaml (if missing)
First, detect the agent name:
bash
# Try to get from existing template.yaml grep "^name:" template.yaml 2>/dev/null | cut -d: -f2 | tr -d ' ' # Or use directory name basename "$(pwd)"
yaml
name: [agent-name-from-directory]
display_name: [Agent Display Name]
description: |
[To be filled - describe this agent's purpose]
resources:
cpu: "2"
memory: "4g"
credentials:
mcp_servers: {}
env_file: []
4c. Create .env.example (if missing)
code
# Environment variables for [agent-name] # Copy to .env and fill in values
4d. Create/Update .gitignore
Ensure these exclusions exist:
gitignore
# Credentials - never commit .mcp.json .env *.pem *.key # Claude Code internals .claude/projects/ .claude/statsig/ .claude/todos/ .claude/debug/ # Runtime/generated content/ session-files/
4e. Create .mcp.json.template (if .mcp.json exists)
Convert .mcp.json to .mcp.json.template with ${VAR_NAME} placeholders:
- •Read
.mcp.json - •Identify credential values (API keys, tokens, paths with user-specific content)
- •Replace with
${VAR_NAME}placeholders - •Write to
.mcp.json.template - •Document variables in
.env.example
Step 5: Create Initial Policy
Create .claude/skills/policy-agent-purpose/SKILL.md:
markdown
--- name: policy-agent-purpose description: Defines this agent's core purpose, scope, and boundaries. user-invocable: false metadata: version: "1.0" created: [today] --- # Agent Purpose Policy ## Identity - **Name**: [agent-name] - **Owner**: [user-name] - **Domain**: [business domain] ## Purpose [What this agent does] ## Scope - **In scope**: [what this agent handles] - **Out of scope**: [what this agent should NOT do] ## Boundaries - Never access systems outside defined integrations - Always confirm destructive operations - Escalate to human when uncertain
Step 6: Confirm Completion
code
## Trinity Adoption Complete ### Created/Updated - [x] template.yaml - [x] .env.example - [x] .gitignore - [x] policy-agent-purpose ### Directory Structure - [x] outputs/ - [x] scripts/ - [x] .claude/skills/ - [x] .claude/agents/ - [x] memory/ ### Next Steps 1. Edit policy-agent-purpose to define this agent's purpose 2. Review and customize template.yaml 3. Run `/trinity-sync push` to deploy to remote 4. Configure notifications with `/trinity-remote notify`
Trinity Requirements Reference
Required Files
| File | Purpose |
|---|---|
CLAUDE.md | Agent's main instructions (the "brain") |
template.yaml | Trinity deployment metadata |
.mcp.json.template | MCP config with ${VAR} placeholders |
.env.example | Documents required environment variables |
.gitignore | Security-critical exclusions |
Required Directories
| Directory | Purpose | Commit? |
|---|---|---|
.claude/skills/ | Agent capabilities | YES |
.claude/agents/ | Sub-agent definitions | YES |
memory/ | Persistent state, schedules | YES |
scripts/ | Automation scripts | YES |
outputs/ | Smaller deliverables | YES |
content/ | Large generated content | NO |
session-files/ | Session-specific work | NO |
.gitignore Must Exclude
Credentials (never commit):
- •
.mcp.json - •
.env - •
*.pem,*.key
Runtime state (discard):
- •
.claude/projects/ - •
.claude/statsig/ - •
.claude/todos/ - •
.claude/debug/ - •
content/ - •
session-files/
Error Handling
| Error | Resolution |
|---|---|
| No CLAUDE.md | Create minimal CLAUDE.md from template.yaml description |
| .mcp.json has secrets | Extract to .mcp.json.template with placeholders |
| Directory not a git repo | Run git init first |
See Also
- •
/trinity-compatibility- Read-only audit - •
/trinity-sync- Git synchronization - •
/trinity-remote- Remote operations - •
/trinity-schedules- Scheduled tasks