tf-compound-patterns

name	tf-compound-patterns
description	Pattern extraction heuristics, pitfall recording format, and compound memory directory structure for capturing learnings from Terraform workflows. Preloaded by all compound-* agents.

Compound Engineering Patterns for Terraform

Core Principle

Each completed workflow run should leave the system better prepared for the next. Knowledge compounds — small, consistent investments in knowledge capture accumulate into exponential productivity gains.

Memory Directory Structure

.foundations/memory/
├── patterns/
│   ├── modules/        # Which module combinations work well together
│   └── architecture/   # Proven architecture templates
├── pitfalls/           # Common mistakes and their resolutions
└── reviews/            # Recurring review issues across runs

Pattern Extraction Heuristics

When extracting patterns from a completed workflow:

1. Module Combinations: Which modules were used together? What inputs connected them?
2. Architecture Decisions: What structure was chosen? Why? What alternatives were considered?
3. Variable Patterns: What validation rules proved useful? What defaults worked well?
4. Security Patterns: What security controls were applied? What was the reviewer's assessment?
5. Testing Patterns: What test strategies worked? What edge cases were discovered?

Pattern File Format

# Pattern: [Descriptive Name]

**Created**: [Date]
**Feature**: [Feature branch/name]
**Confidence**: [High|Medium|Low] (based on review scores and deployment success)

## Context
[When to use this pattern]

## Module Combination
[List of modules used together with version constraints]

## Key Configuration
[Important settings and their rationale]

## Lessons Learned
[What worked well, what to watch out for]

Pitfall Recording Format

# Pitfall: [Short Description]

**Discovered**: [Date]
**Feature**: [Feature branch/name]
**Severity**: [Critical|High|Medium|Low]
**Phase**: [Planning|Implementation|Testing|Deployment]

## Symptoms
[What went wrong — observable behavior]

## Root Cause
[Why it happened]

## Resolution
[How it was fixed]

## Prevention
[How to avoid this in future — checklist items, validation rules, etc.]

## Related
[Links to patterns, other pitfalls, or documentation]

AGENTS.md Update Patterns

When updating AGENTS.md files:

• Add implementation details discovered during the workflow
• Document debugging insights and architectural patterns
• Record dependency relationships between components
• Note common pitfalls specific to the component area
• Keep updates minimal and high-signal — no noise

Constitution Review Patterns

When reviewing constitution alignment:

• Identify gaps where the constitution didn't cover a scenario
• Note principles that were too strict or too loose
• Suggest new MUST/SHOULD/MAY rules based on experience
• Document exceptions that were needed and why

Template Improvement Patterns

When assessing templates:

• Identify sections consistently skipped → consider making optional
• Identify information consistently added → consider making a section
• Note formatting issues that caused friction
• Suggest reordering based on actual workflow

Compound Phase Triggers

Run outcome is determined by deploy_status from tf-implement Step 4a:

• RUN_SUCCESS = (deploy_status == "success")

Agent	Runs when	Receives
compound-pattern-extractor	RUN_SUCCESS only	deploy_status, feature dir, report path
compound-pitfall-recorder	Always	deploy_status, feature dir, failure: true if !RUN_SUCCESS
compound-agents-updater	RUN_SUCCESS only	feature dir
compound-constitution-reviewer	Review flagged constitution issues	feature dir, review findings
compound-template-improver	Template deviations detected	feature dir, template paths

• Auto-commit: bash .foundations/scripts/bash/checkpoint-commit.sh --dir .foundations/memory/ --prefix compound "learnings"
• Best-effort: Compound failures don't block the workflow