Adinkra Labs — SAGE Operations Architecture

The Two-Tier
Execution Model

SAGE agents do the cheap prep work. Claude Code does the expensive execution. Every Opus token goes toward building, not orienting.

4
Autonomous Agents
$0.25
Per Prep Task
2-3x
Productivity Gain
90%+
Execution Context

The Orientation Tax

Claude Code (Opus 4.6, 1M context) can hold entire codebases and execute complex multi-file changes in one shot. But without pre-structured work, sessions spend 30-50% of their expensive context just figuring out what to do.

Activity
% of Opus Context
Cost Impact
Reading trackers, understanding state
30-50%
$2-8 per session wasted on orientation
Research — find files, read code, check deps
15-25%
Haiku-level work at Opus prices
Task decomposition
5-10%
Sonnet can do this for $0.25
Actual execution
20-50%
Half or less of your budget builds anything

Two Tiers, One Pipeline

Route each activity to the cheapest model that can handle it. SAGE agents run continuously on Sonnet/Haiku. Claude Code fires on-demand for complex execution.

Tier 1: SAGE Agents
Tier 2: Claude Code
Model Class
Sonnet / Haiku
Opus 4.6 (1M context)
Cost
$0.25-$0.80 per task
$2-$8 per task
Runs
Continuously (heartbeats)
On-demand (sessions)
Does
Research, context gathering, task decomposition, priority scoring, dependency mapping, status monitoring
Complex code execution, multi-file changes, architecture decisions, full build sessions, creative problem solving
Produces
SAGE Execution Prompts (ready-to-paste)
Shipped code, PRs, deployed features
The Key Insight

The handoff mechanism is a self-contained Execution Prompt — a structured, ready-to-paste instruction block placed in a Handoff Queue. Claude Code picks it up and executes immediately. No research phase needed. 90-100% of context goes toward building.


Four Specialists, One Pipeline

Each SAGE agent owns a slice of the prep work. Together they produce execution-ready work packages that Claude Code can act on without orientation.

Prepares Execution Prompts

Project Manager

Scans trackers, identifies highest-priority work, gathers all context (files, deps, Figma nodes), decomposes tasks into atomic steps, writes structured Execution Prompts.

Maintains Quality

Process Auditor

Audits session logs, backlog items, Notion organization. Flags stale items, duplicates, and orphan pages. Produces actionable remediation lists.

Guards Infrastructure

DevOps Monitor

Checks deployment health across all services. Flags incidents with severity and recommended action. Produces incident reports that escalate to execution prompts when code fixes are needed.

Orchestrates

Ops Director

Reviews specialist outputs, resolves priority conflicts across projects, approves escalations to Claude Code, and produces weekly operational summaries.


The Contract

Every task flowing from SAGE to Claude Code must be a self-contained instruction block. No additional research needed.

## SAGE Execution Prompt — [SAG-XX]

### Objective
[One sentence: what needs to happen]

### Context
- Project: [Atlas / OpenClaw / Delphi OS / KAA]
- Lane: [PORTAL / BACKEND / CODEX / OPENCLAW / INFRA]
- Source: [Build Tracker BO XX / Backlog ID / Incident]
- Priority: [Critical / High / Medium]
- Dependencies: [What must be true before starting]

### Files of Interest
- path/to/file1.ts — [why this file matters]
- path/to/file2.tsx — [what needs to change]

### Execution Steps
1. [Specific, imperative step]
2. [Next step]
3. [...]

### Acceptance Criteria
- [ ] [Measurable outcome]
- [ ] next build passes with 0 errors

### References
- Figma: [node ID if UI work]
- Related: [SAG-XX, Backlog ID]

From Prep to Execution

A continuous pipeline where SAGE agents prep and Claude Code executes — no gaps, no wasted context.

Step 01

SAGE Agent Completes Prep

Project Manager structures the work, gathers all context, writes a self-contained Execution Prompt with files, steps, and acceptance criteria.

Step 02

Creates Handoff Queue Packet

Execution Prompt placed in the Handoff Queue with Status: Pending, Direction: SAGE to CC, priority level, and target session lane.

Step 03

Claude Code Session Starts

Session Start auto-behavior checks the Handoff Queue. Finds Pending packet, claims it, reads the Execution Prompt.

Step 04

Immediate Execution

No research phase needed. 90-100% of context goes toward building. Multi-file changes, architecture decisions, full build sessions.

Step 05

Completion & Feedback

Post-build verification, session log with resume point, tracker updates. SAGE agents pick up completed status and prep the next task.


The Numbers

The same session budget, dramatically more productive. SAGE pre-digests everything so Opus only does what Opus is for.

Activity
Without SAGE
With SAGE
Orientation (read trackers, understand state)
30-50% of Opus context
0% — SAGE pre-digested
Research (find files, read code, check deps)
15-25% of Opus context
0% — SAGE mapped it
Task decomposition
5-10% of Opus context
0% — SAGE structured it
Actual execution
20-50% of Opus context
90-100%
Net Effect

2-3x more productive Opus sessions at the same token cost. Every dollar spent on SAGE prep ($0.25-$0.80) saves $2-8 of Opus context that would have been wasted on orientation. The ROI compounds with every session.