Guide

Advanced

Signature-Based Locking

REAP uses a cryptographic nonce chain to enforce stage ordering. Without a valid nonce, the AI agent cannot advance to the next stage — even if it tries to skip ahead.

text

Stage Command          current.yml              /reap.next
─────────────          ───────────              ──────────
generate nonce ──────→ store hash(nonce)
return nonce to AI                         ←── AI passes nonce
                                               verify hash(nonce)
                                               ✓ advance stage

How It Works

Stage command (e.g. /reap.objective) generates a random nonce
The nonce's SHA-256 hash is stored in current.yml
The nonce is returned to the AI agent in the JSON response
/reap.next receives the nonce, hashes it, and compares against current.yml
Match → stage advances. Mismatch → rejected.

Prompt-Only vs Signature-Based

Threat	Prompt-Only	Signature-Based
Skipping stages	Relies on AI compliance	Blocked — no valid nonce
Forging tokens	N/A	Infeasible — one-way hash
Replaying old nonces	N/A	Blocked — single-use, stage-bound
Prompt injection	Vulnerable	Nonce is external to prompt context

Lineage Compression

As generations accumulate, lineage archives are automatically compressed during the Completion stage.

Level	Input	Output	Trigger	Protection
Level 1	Generation folder (5 artifacts)	gen-XXX-{hash}.md (40 lines)	lineage > 5,000 lines + 5+ generations	Recent 3 + DAG leaf nodes
Level 2	100+ Level 1 files	Single epoch.md	Level 1 files > 100	Recent 9 + fork points

DAG preservation: Level 1 files retain metadata in frontmatter. Level 2 epoch.md stores a generations hash chain. Fork guard: all local/remote branches are scanned before Level 2 compression — fork points are protected. Epoch-compressed generations cannot be used as merge bases.

Entry Modes

Specified with reap init --mode. Controls how the Genome is initially structured.

Mode	Description
greenfield	Build a new project from scratch. Default mode. Genome starts empty and grows.
migration	Build anew while referencing an existing system. Genome is seeded with analysis of the existing system.
adoption	Apply REAP to an existing codebase. Genome starts from templates and is populated during the first generation's Learning stage.