Borrowing it
Nothing to install: this file belongs to nodealchemy/powernode-platform. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/nodealchemy/powernode-platform/develop/.claude/skills/dev-loop/SKILL.mdgit clone --depth 1 https://github.com/nodealchemy/powernode-platformWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nodealchemy/powernode-platform/dev-loop)<a href="https://agentmods.dev/skills/nodealchemy/powernode-platform/dev-loop"><img src="https://agentmods.dev/badge/skills/nodealchemy/powernode-platform/dev-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nodealchemy/powernode-platform/dev-loop"><img src="https://agentmods.dev/badge/skills/nodealchemy/powernode-platform/dev-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.01884 |
| Opus 5.5 | $0.00000 | $0.00754 |
| Sonnet 5.5 | $0.00000 | $0.00377 |
| Haiku 4.5 | $0.00000 | $0.00188 |
Grade A, and why
dev-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 111 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/dev-loop — Execute One Dev-Loop Iteration
Pull the next task from a platform Ralph Loop queue (via the dev_loop MCP bridge) and drive it to a verified, committed, reported outcome. One task per invocation: a Ralph-pattern loop body, driven repeatedly by /loop /dev-loop.
Usage
/dev-loop # default loop: dev-improve
/dev-loop <loop-name-or-id> # any other claude_code Ralph Loop
Delegate every task to a subagent (REQUIRED)
The session that runs /dev-loop is the driver. It pulls and routes; it does not implement. Steps 2–8 below run inside a subagent:
- Driver does: step 1 (Pull); picking the worktree (step 3, including a liveness check that no other session owns it); spawning the executor; relaying its outcome to the operator; scheduling the next iteration.
- Executor:
platform-engineerfor code, spec and doc changes (the default: it has Edit/Write/Bash and loads the dev-loop verbs on demand),fleet-operatorfor a task that acts on the running fleet. Useplatform.route_taskwhen the area is unclear; while the legacy canonicals still run the platform (C4 shadow mode) it may name one of them, so spawn the roster agent for that area. Brief it with the full task (key, acceptance criteria,metadata.files, guardrails), the worktree path and branch, and the red-first / commit /dev_complete_taskcontract: a subagent starts without this session's context. The executor owns steps 2–8 and reports the task itself. - Reviewer: the canonical executor agents (e.g.
platform-engineer) have no Agent tool, so they cannot spawn a reviewer. When acceptance criteria require independent review, the executor stops before committing and reports its diff. The driver then spawnsplatform-reviewer— read-only, with no shell in phase 1 (no Bash, Edit or Write) — and passes it the diff and the worktree path to read; the driver runs any git or test command the review names. The driver sends the findings back to the same executor (SendMessage), which fixes them, commits, and callsdev_complete_task. An executor must never replace this review with a self-review. - Blocked or ambiguous tasks: the driver takes the decision to the operator (AskUserQuestion, one decision at a time) and records it with
dev_update_task/dev_complete_task. The executor does not guess at operator decisions. - One executor per worktree. Never pull the next task onto a worktree whose executor is still running; use another, within the
WORKTREE_MAXcap below.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago Changed · -1 lines e50ecdf392d9
- 8d ago Changed ec0ec2e7161c
- 11d ago Changed 9920446be78b
- 1mo ago Changed c6c29d65e202
- 1mo ago Changed · +10 lines a63c730c1749
- 2mo ago First seen · 102 lines · 0 tokens per session scan A 486624284170
dev-loop is a skill published in the GitHub repository nodealchemy/powernode-platform (4 stars, last pushed 5d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,884 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
gitlab_read
Read access: projects, issues, MRs, files, pipelines, members.
epic-close
Close out a completed EPIC / sprint end-to-end: author a fact-verified meta-post announcing what shipped, archive the EPIC's handovers, snapshot orchestrator memory, gate the integration→main merge on CI-green + explicit human go, drive the version bump (merge-commit conventions that auto-sync classifies), and verify…
gh-assign-issues
Use to assign GitHub issues to a milestone and/or owners in bulk, verifying each.
orchestrate
Toolbox for driving HumanLayer from the CLI - log in, pick a host, launch tasks in their own worktrees, launch sessions, list tasks and their PRs, and copy, move, or upload task artifacts.
meta-issue-to-pr-autopilot
Triages an issue, delegates the fix to sub-agent, drafts a PR description with summarize, and opens the PR via gh. Best used on small, well-scoped issues with clear acceptance criteria.
meta-github-pr-watch-digest
Inspect the user's open GitHub PRs / failing CI / new issues via gh, summarize into 3 buckets (to-review / awaiting-me / CI-red), and persist follow-ups to memory.