eli-labz/godcoder/loop-engineering-bruteforce

Loop Engineering Bruteforce Operator

Use this skill when you need to install and operationalize the local loop stack in this repo: third_party/loop-engineering-main + third_party/ResearchSwarm-master.

Goal

Drive repeatable autonomous improvement loops in the GodCoder codebase by combining:

• Loop design, scaffolding, audit, and cost tooling from loop-engineering.
• Persistent route/log/optimize feedback via the ResearchSwarm bridge.

Baseline bootstrap (Windows)

From repo root:

tools\bootstrap-third-party-stack.bat

This installs dependencies for both third-party projects and validates:

• godcoder_harness.py recall --limit 1
• node tools/loop-sync/dist/cli.js .

Bruteforce loop workflow

1. Route work through ResearchSwarm first:

py third_party/ResearchSwarm-master/godcoder_harness.py route "<objective>"

1. Initialize or refresh loop artifacts in the target folder:

cd third_party/loop-engineering-main
npx @cobusgreyling/loop-init .. --pattern daily-triage --tool codex

1. Audit readiness and collect fixes:

npx @cobusgreyling/loop-audit .. --suggest

1. Estimate token cost before escalating cadence:

npx @cobusgreyling/loop-cost --pattern daily-triage --level L1

1. Log concrete outcomes back to memory with stable tags:

py third_party/ResearchSwarm-master/godcoder_harness.py log --status success --tag loop:daily-triage --instruction "<objective>" --summary "<one-line result>"

1. Re-rank approaches and prefer the highest-yield tags:

py third_party/ResearchSwarm-master/godcoder_harness.py optimize

Operating rules

• Keep one measurable change per iteration.
• Verify with build/test/audit before logging success.
• Reuse stable tags (loop:<pattern>) so optimize output remains sharp.
• For risky or ambiguous operations, keep the loop in report-only mode first.