Skip to main content

​
Summary

Built the FORGE Deep Research (DR) Handbook — a multi-repo technical handbook that enables fully autonomous agent-to-agent (A2A) software development lifecycle execution. The handbook replaces manual “GO TASKSET N” gating with zero-human-intervention overnight execution via n8n FORGE.

​
What changed operationally

​
New Agent Suite in Veritas

A new deep-research/ agent suite was added to veritas/v3/agents/ with two XML prompt specifications (FORGE-DR-001 for state auditing, FORGE-EX-001 for execution planning) plus an orchestrator, README, and playbook. Action needed: Register these prompts with any downstream FORGE automation that consumes the Veritas agent registry.

​
Prompt Library Registry Updated

Two new entries added to veritas/prompts/index.json (vrt-f0d3e8a1, vrt-b7c9d2f4), bringing total reviewed prompts from 114 to 116. Config change: No config changes needed — index.json is consumed dynamically.

​
4-Stream Verification Pattern Formalized

The FORGE Code 06 (Churnery) verification pattern was formalized as a reusable XML element (<verification_streams>) embedded in both prompt specifications. This can be extracted as a Veritas fragment prompt for reuse. Action needed: Consider creating a fragment/verification-4-stream prompt for reuse across future FORGE prompts.

​
Autonomous Overnight Execution Protocol

The execution manifest (execution_manifest.yaml) is designed for direct POST to n8n FORGE webhook. Throttling parameters (3 concurrent, 30 req/min, 60s cooldown) are embedded in the manifest schema. Action needed: Implement the n8n FORGE webhook endpoint that accepts execution manifests if it doesn’t already exist.

​
Business impact

  • Eliminates overnight blocking on manual taskset approval — previously required human “GO” between each taskset
  • Enables continuous overnight execution of development work across 4 organizations and 15+ repos
  • Provides systematic state auditing that catches anomalies (hollow bundle completions, branch drift, duplicate PRs) that manual reviews miss
  • Formalizes the A2A handoff protocol with retry semantics, enabling graceful degradation on failure
  • Establishes a reusable 4-stream verification pattern that raises confidence in AI-generated analysis