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 instructions/ericyuan2007/gpuplane/agents-mdgit clone --depth 1 https://github.com/EricYuan2007/GPUPlaneWhat 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 | $0.01751 | $0.01751 |
| Opus 5 | $0.00875 | $0.00875 |
| Sonnet 5 | $0.00350 | $0.00350 |
| Haiku 4.5 | $0.00175 | $0.00175 |
Grade A, and why
GPUPlane AGENTS.md 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- GPUPlane CLAUDE.md — 92% identical, 6 lines differ
How it starts
The opening of the file, as written. The whole thing — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — GPUPlane agent guide
Canonical agent guidance for this repo.
CLAUDE.mdmirrors it; keep both in sync when editing. Load this before working in GPUPlane.
GPUPlane is an agent-native training control plane for personal GPUs. This
file is the short rulebook an AI agent (Claude Code, Codex, Cursor, …) follows
when running or extending it. The full design lives in docs/.
What this project is (and isn't)
- Is: a control plane that schedules training jobs on a single workstation / small GPU fleet, observes them with training semantics (run / checkpoint / evaluation), and exposes a read+write MCP surface so an agent can drive the experiment loop.
- Isn't: a W&B/MLflow replacement (no experiment tracking ambition), a Docker/K8s orchestrator (jobs are bare processes on the agent host), or a multi-tenant cloud product.
Stay on the bare-process + SQLite side of the line. Do not add containers, queues-as-a-service, or cloud SDKs to satisfy a request that "would be nicer in the cloud" — the whole value proposition is low-infra local GPU.
The experiment loop
run-experiment → monitor-experiment → analyze-results → (keep / revert)
submit_job get_run_summary compare_runs checkpoint path
list_events get_best_checkpoint git keep/revert
diagnose_run evaluate_checkpoint
The three skills in .claude/skills/gpu-training/ encode this loop. Prefer
invoking them over improvising with the MCP tools directly — they carry the
conventions below.
Conventions
Metrics
- Training scripts must log metrics via the GPUPlane SDK or TensorBoard.
Without metrics,
diagnose_run/ anomaly detection / recommendation have nothing to read. Metric names are namespaced (train/loss,eval/loss). - One primary metric drives every "best" decision. It resolves through
experiment → project → global inheritance; cross-scope comparisons use only
their common scope or an explicit metric. Default
eval/loss(minimize) oreval/accuracy(maximize). Don't multi-objective.
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 · 145 lines · 1,751 tokens per session scan A 587db89dc657
GPUPlane AGENTS.md is an instructions file published in the GitHub repository EricYuan2007/GPUPlane (0 stars, last pushed 7d ago), licensed Apache-2.0. It adds 1,751 tokens to every session, about $0.0088 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 instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.