Jeden Skill in Manus ausführen
mit einem Klick

Jeden Skill in Manus mit einem Klick ausführen

heavy-coding-eval

Sterne21

Forks1

Aktualisiert25. Juni 2026 um 06:18

Design and scaffold evaluations comparing single-candidate Composer coding against adaptive Heavy Coder candidate teams.

Installation

Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.

In Manus ausführen

Quelle

codegraphtheory

codegraphtheory/heavy-coder

GitHub-Repository öffnen Creator-Repositorys ansehen

Download

In Manus ausführen

Verwandte BerufeSOC

Basierend auf der SOC-Berufsklassifikation

SoftwareentwicklerInformatik- und Mathematikberufe·SOC 15-1252

Datei-Explorer

5 Dateien

SKILL.md

readonly

name	heavy-coding-eval
description	Design and scaffold evaluations comparing single-candidate Composer coding against adaptive Heavy Coder candidate teams.
version	0.1.0
author	CodeGraphTheory
license	MIT

Heavy Coding Eval

Use this skill when planning or running Heavy Coder evaluations.

Current status

Scaffolded. Benchmark execution is not implemented.

Protocol

Control: one Composer candidate, same outer coordinator and verifier, no comparative critic.
Treatment: adaptive 3/5 Composer candidates (per profile candidate_widths), blind comparative critic, reasoning-model synthesis when available, same verifier.
Initial target: preregistered 10 to 20 task subset, preferably SWE-Bench Pro Public if current licensing permits.
Three runs per task per condition when feasible.
Primary endpoint: official benchmark resolution.
Report uncertainty, model-call count, cost, and wall-clock time.
Never present a subset as a full leaderboard score.

Mehr aus diesem Repository

gleiches Repository

heavy-context-budget

codegraphtheory/heavy-coder

Keep coordinator and leaf context lean-batch reads, cap logs, use execute_code to filter-so swarms stay fast and within token limits.

2026-06-2521

heavy-explore-first

codegraphtheory/heavy-coder

Use at the start of non-trivial software tasks. Gather repo ground truth with batched reads and searches before planning or delegate_task.

2026-06-2521

heavy-hook-phases

codegraphtheory/heavy-coder

Obey Plan 1A hook session phases so mutating tools, delegate_task sizing, and re-planning do not waste turns or trigger blocks.

2026-06-2521

heavy-leaf-brief

codegraphtheory/heavy-coder

Use when spawning delegate_task leaves for coding work. Pack self-contained goals and context so parallel Composer workers ship verifiable patches without coordinator chat history.

2026-06-2521

heavy-leaf-candidate-output

codegraphtheory/heavy-coder

Require structured candidate-result JSON from swarm leaves so critique_candidates.py and synthesis run on evidence, not prose self-reports.

2026-06-2521

heavy-pre-dispatch-enrich

codegraphtheory/heavy-coder

Enrich slim hook-injected delegate_tasks with touch map, failure excerpts, and verification commands so parallel leaves produce verifiable patches without coordinator chat history.

2026-06-2521

name	heavy-coding-eval
description	Design and scaffold evaluations comparing single-candidate Composer coding against adaptive Heavy Coder candidate teams.
version	0.1.0
author	CodeGraphTheory
license	MIT

Heavy Coding Eval

Use this skill when planning or running Heavy Coder evaluations.

Current status

Scaffolded. Benchmark execution is not implemented.

Protocol

Control: one Composer candidate, same outer coordinator and verifier, no comparative critic.
Treatment: adaptive 3/5 Composer candidates (per profile candidate_widths), blind comparative critic, reasoning-model synthesis when available, same verifier.
Initial target: preregistered 10 to 20 task subset, preferably SWE-Bench Pro Public if current licensing permits.
Three runs per task per condition when feasible.
Primary endpoint: official benchmark resolution.
Report uncertainty, model-call count, cost, and wall-clock time.
Never present a subset as a full leaderboard score.