| name | loop-engineering-bruteforce |
| description | Use the vendored loop-engineering toolkit to scaffold, audit, and continuously improve autonomous loops in this GodCoder repository, and pair the result with the ResearchSwarm bridge for route/log/optimize memory. |
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 1node tools/loop-sync/dist/cli.js .
Bruteforce loop workflow
- Route work through ResearchSwarm first:
py third_party/ResearchSwarm-master/godcoder_harness.py route "<objective>"
- Initialize or refresh loop artifacts in the target folder:
cd third_party/loop-engineering-main
npx @cobusgreyling/loop-init .. --pattern daily-triage --tool codex
- Audit readiness and collect fixes:
npx @cobusgreyling/loop-audit .. --suggest
- Estimate token cost before escalating cadence:
npx @cobusgreyling/loop-cost --pattern daily-triage --level L1
- 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>"
- 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.
