Borrowing it
Nothing to install: this file belongs to Nauro-AI/nauro. 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/Nauro-AI/nauro/main/.agents/skills/nauro-loop/SKILL.mdgit clone --depth 1 https://github.com/Nauro-AI/nauroWrote 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/nauro-ai/nauro/nauro-loop)<a href="https://agentmods.dev/skills/nauro-ai/nauro/nauro-loop"><img src="https://agentmods.dev/badge/skills/nauro-ai/nauro/nauro-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/nauro-ai/nauro/nauro-loop"><img src="https://agentmods.dev/badge/skills/nauro-ai/nauro/nauro-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 | $0.00184 | $0.04858 |
| Opus 5 | $0.00092 | $0.02429 |
| Sonnet 5 | $0.00037 | $0.00972 |
| Haiku 4.5 | $0.00018 | $0.00486 |
Grade A, and why
nauro-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 10d 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 — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Nauro loop skill
Run gated work origination and coordinate selected Program Delivery. The Program state machine is FRAME -> CHOOSE -> START -> ADVISE -> VERIFY -> ADVANCE. The coordinator owns the verified frame, sequence, handoff, advisory review, integration verification, and next recommendation. It does not implement, approve, file project truth, push, create a pull request, merge, or start the next slice.
User-named work enters as HUMAN-SELECTED. Agent-originated work keeps read-only ORIENT, 1-3 ranked Delivery and Interview candidates, mandatory SELECT, reject all, and no auto-pick path. Interview remains explicit, separate, and non-authoritative. Synchronous non-program Delivery stays outside the Program state machine and keeps the current nauro-ship-task chain.
Authority boundary
These are authority rules. MCP or shell write paths may exist, but they do not grant the coordinator authority to use them for Program Delivery.
- The loop cannot automatically change project truth or file decisions. Ordinary prompts, Interview results, Delivery handoffs, and program handbacks create no automatic store artifacts. The scheduled ORIENT selection-checkpoint writes are the narrow exception: they create only the existing SELECT checkpoint and pointer through the filesystem and
nauro syncmechanism defined below. A selected Interview returns candidate shared understanding only. After the interview, an explicitly approvedpropose_decision,update_state, orflag_questionpayload may be executed through/nauro-interviewaccording to its contract and separate later approval gates. Those writes belong to the interview contract, not loop authority. - The coordinator may inspect pull-request and merge state through read-only channels, including
gh pr viewor an equivalent authenticated read channel, only when a required PR-backed repository-anchor check needs it. This includes ORIENT, Resume, replacement-coordinator recovery, and VERIFY. Read-only inspection does not grant publication authority. The coordinator never pushes, creates or edits a pull request, merges, or performs another publication mutation. Only the direct-user Delivery task may perform approved publication work through its own gates. - The loop is NOT a "keep moving" override of any inner gate. A standing "keep going" or auto-mode directive does not clear the SELECT gate, the plan gate, a tech-lead pause, or the push gate. The loop exists to repeat the gated chain, not to bypass it.
- SELECT is never auto-picked. Neither entry mode picks for the human, not even when exactly one candidate ranks or when a scheduled continuation has one surviving candidate. The synchronous mode surfaces SELECT in the parent session; the scheduled mode parks a SELECT checkpoint and exits before any gate; the resume continuation surfaces SELECT to the human. No path resolves SELECT without the human.
- A hidden child, ordinary subagent, generic agent, or persistent Delivery child is not a substitute for a fresh direct-user Delivery task.
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.
- 10d ago First seen · 166 lines · 184 tokens per session scan A 3f87beaf46c3
nauro-loop is a skill published in the GitHub repository Nauro-AI/nauro (10 stars, last pushed yesterday), licensed Apache-2.0. It adds 184 tokens to every session and 4,858 once invoked, about $0.0009 per session on Opus 5. 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
design-inventory
Use to run the Claude Design to ClosedLoop pipeline against the current web-ui. Stage A inventories a design export zip into schema-validated findings (typed design units - screens, regions like nav bars, standalone components like a chat dialog; UX and behavioral changes; Storybook component reuse mapping; token…
prd-creator
This skill helps PMs draft lightweight PRDs. It should be used when a PM wants to define a new feature, brainstorm requirements, or prep for sprint planning. Triggers include "I have a feature idea", "help me write a PRD", "let's break this into stories", or "what do I need for sprint planning". Optimized for speed …
upload-artifact
Upload a file as a ClosedLoop document (PRD, implementation plan, feature, or template). Reads file content and uploads via MCP without consuming conversation context. Also supports creating new versions of existing documents. Triggers on: "upload artifact", "upload PRD", "upload implementation plan", "upload…
cross-repo-cache
Check if cross-repo coordinator results can be reused, avoiding redundant Sonnet agent launches. Compares peer repo git hashes against stored hashes from last coordinator run. Triggers on: entering Phase 1.4.1, checking cross-repo cache, before discovering peers. Returns CROSSREPOCACHEHIT with cached status or…
execute-plan
Implement a written plan (from write-plan or interview-plan) when the user says execute or implement it. Loads the plan, executes ready tasks in waves in the ticket venue, runs the plan's review gate itself. Do not use while still interviewing or writing the plan.
lithium-kb
Generates, maintains, and visualizes a structured Markdown knowledge base (.lithium-kb/{architecture,debug,tasks,features}) with an interactive Neural Network graph, auto-watch mode, custom agent directives (.agentrules), and MCP server integration for Pi, Claude, Codex, Cursor, and other coding agents. Triggers…