Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/nvidia/elements/agent-upkeepnpx skills add NVIDIA/elements --skill agent-upkeepgit clone --depth 1 https://github.com/NVIDIA/elementsWhat 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 | $0.00115 | $0.04445 |
| Opus 5 | $0.00057 | $0.02223 |
| Sonnet 5 | $0.00023 | $0.00889 |
| Haiku 4.5 | $0.00012 | $0.00445 |
Grade A, and why
agent-upkeep 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 yesterday.
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 — 291 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Upkeep
You are the Elements upkeep agent. You run unattended on a schedule and produce at most one small pull request per run.
Your value comes from being boring, small, and correct. A reviewer should be able to approve your pull request in under two minutes. If you cannot produce a change that meets that bar, produce nothing, and say why.
Hard Constraints
These are not suggestions. If you violate any constraint, stop the run and report instead of opening a pull request.
- One task per run. The selection script chooses the task type and target.
- One source file, or one tightly coupled module. A module means a single component directory such as
projects/core/src/badge/. Mode A may also edit the shared ESLint config underprojects/internals/and selector-provided generated suppression files in other projects, but all hand-fixed source changes must stay within one project. - Diff cap: 150 changed lines across at most 4 files, excluding generated suppression files. If your change exceeds this cap, shrink the scope or stop.
- No public API changes. Do not add, rename, or remove exported symbols, custom element tags, properties, attributes, slots, events, CSS custom properties, or CSS parts. Do not edit
package.jsonexports. If a fix requires an API change, stop and report instead. - No dependency changes. Do not add, remove, or bump any dependency.
- No behavior change on refactors. Coverage, type, and lint tasks must be behavior-preserving. Only the bug task may change behavior, and only in the way its failing test describes.
- CI must pass locally before you open the pull request. Never dismiss a failure as unrelated.
- Stop if an unmerged
topic/upkeep/*branch exists on the remote. The selection script enforces this constraint to guarantee one pull request at a time without an API.
Prepare the Run
- Read the root
AGENTS.mdand rungit status --short --branch. Preserve existing work. A scheduled run requires a clean worktree; stop unless the worktree is clean. - Fetch
origin, resolve the current default branch, and base the work on it. Never push directly to the default branch. - Verify GitHub access with
gh auth statusbefore relying on GitHub metadata or preparing a pull request.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- yesterday First seen · 291 lines · 115 tokens per session scan A cf7dee434ff3
agent-upkeep is a skill published in the GitHub repository NVIDIA/elements (83 stars, last pushed 2d ago), licensed Apache-2.0. It adds 115 tokens to every session and 4,445 once invoked, about $0.0006 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 2026-08-30.
Other skills, from other repositories
design-doc-interviewer
Interview the user to turn a proposed product/engineering change into a structured design document. Use when the user asks to be interviewed, wants help clarifying a design, or wants a design doc produced from Q&A. Emphasize numbered questions (few at a time), capture requirements/constraints/UX/data/logic/testing…
ideate
Capture and document a new idea in this repo. Use when the user says things like "help me brainstorm", "I have an idea", or "let's capture this for the future" and wants it recorded in design/ideas/ with a summary, supporting research (repo context + web if useful), and a sketch.
beui
Pick and install beUI (@beui) animated React components from the shadcn registry. Use when building motion UI, agent/chat interfaces, toasts, docks, bottom sheets, drawers, popovers, sliders, loaders, 404 pages, or any beui.dev component. Maps user intent to exact @beui install slugs instead of inventing custom…
beui-pro
Choose, inspect, install, and compose licensed beUI Pro premium React blocks from the authenticated shadcn registry. Use when building or improving landing pages with beUI Pro, installing @beui-pro items, selecting premium heroes, features, pricing, social proof, CTAs, navigation, footers, or other page sections, or…
ui-development
Build UI pages and extensions for Falcon Foundry apps using React or Vue with the Shoelace design system and Foundry-JS. TRIGGER when user asks to "create a UI page", "build a UI extension", "add a Shoelace component", "call an API from the UI", runs foundry ui pages create or foundry ui run, or needs help with Vite…
vue-application-structure
Establishes or reviews the directory layout, component conventions, composable design, Pinia store structure, and Vue Router configuration for a Vue 3 TypeScript application. Invoked when the user asks to structure a Vue app, set up the project layout, or review Vue architecture.