StackMap
Subscribe

bemyagent vs loop-engineering

One markdown bootstrap file that scaffolds a structured workspace for any coding agent — persistent project docs plus a task tree run through a Think→Task→Execute→Verify cycle with human pacing gates. — versus — Reference repo plus npm CLIs (loop-init/audit/cost) for loop engineering: designing scheduled, gated control loops that prompt and orchestrate AI coding agents — Grok, Claude Code, Codex — over time.

The curated verdict

Both impose a disciplined, tool-agnostic operating structure on coding agents through scaffolding, an explicit work cycle and human gates. bemyagent structures a single human-paced session (Think→Task→Execute→Verify); loop-engineering structures scheduled, unattended loops that run over time. Pick bemyagent for interactive pacing, loop-engineering for autonomous automation.

bemyagentloop-engineering
Stars228.0k
Forks61.1k
LanguageHTMLJavaScript
LicenseMITMIT
Last activity8 days agotoday
Topicscoding, memorycoding, orchestration
Curated connections47

bemyagent — the curator's take

Try it when your agent keeps drifting on multi-session projects and a flat CLAUDE.md isn't enough structure: one file bootstraps a disciplined docs+work tree, it's agent-agnostic (self-registers into .cursorrules/AGENTS.md), and INTERACTIVE mode gives you real plan/result approval gates. NOT for quick scripts or single-session work — the ceremony outweighs the benefit — and it's a young one-maintainer protocol: no enforcement layer exists, so compliance depends entirely on the model faithfully following a long ruleset; weaker models will ignore half of it. Expect to adapt conventions yourself rather than lean on tooling or community.

loop-engineering — the curator's take

The methodology layer for autonomous coding agents — read it once you've stopped hand-prompting and want to design the loop that prompts the agent instead: scheduled triage, worktree isolation, maker/checker sub-agents, MCP connectors, human gates, and a phased L1-report → L3-unattended rollout. Ships 7 clone-and-run patterns, tool-agnostic starters (Grok/Claude Code/Codex/OpenCode), and CLIs that score loop-readiness (loop-audit), estimate token spend (loop-cost) and scaffold state/budget (loop-init). It's a patterns + tooling reference, NOT a runtime — it won't execute or host your agents, so bring your own harness (an alook-style AI-company layer or a squid-style pipeline). Overkill if you just want one-shot interactive help; heed its own caveat that unattended loops make unattended mistakes, so verification stays on you.