Skip to content

Engineering Handbook Integration

Hero

I maintain a personal engineering handbook covering testing, security, source management, observability, configuration, and documentation standards. This update distills those standards into guidelines that architect agents can query via Cognee. I've been working on refining it over the past few days and gave it the good 'ole v1.0.0 tag. Next, I wanted to distill all that into information that could be put into Mnemonic and used by the agents.

RLM Analysis

Note: I've been using Brainqub3's RLM agent and skill for similar purposes in other projects, and man, I'm very happy with it!

I used an RLM approach adapted from Brainqub3 - John Adeojo's - claude_code_RLM project - his video walkthrough explains the technique. I modified it to work under $HOME/.claude because for some reason the rlm subcall agent wasn't "available" according to Claude when defined at the project level. Once moved, it worked perfectly. It compared 14 handbook documents against 57 existing agent patterns. Nice!

Findings

The analysis surfaced:

  • 4 minor conflicts - Different resource threshold recommendations (separate concerns, not contradictions)
  • Major gaps - Handbook coverage for observability, configuration, and documentation had no corresponding patterns
  • Overlaps - E2E testing philosophy, Docker builds, security scanning already aligned
  • Augmentation opportunities - Existing patterns could reference handbook standards for authoritative guidance

Filtering: 16 to 11 Guidelines

The initial analysis produced 16 candidate guidelines. I filtered to 11 development-focused guidelines that directly impact architecture, design, or implementation.

Removed (operational concerns):

  • zero-downtime-guideline - Deployment operations
  • cd-automation-threshold-guideline - Release process decisions
  • rollback-strategy-guideline - Incident response procedures
  • secrets-rotation-guideline - Operational security practices
  • runbook-template-guideline - Operational documentation format

These are valid standards but better suited for runbooks. Architect agents design systems; they don't execute deployments or manage incidents.

The 11 Guidelines

Located in agent-patterns/engineering-guidelines/:

Domain Guidelines
Observability structured-logging, metrics-instrumentation, health-check, distributed-tracing
Configuration config-precedence, startup-validation
Testing coverage-thresholds
Source Management branch-strategy, branch-protection
Documentation readme-structure, changelog