cursor-models

A rule that makes tasks and helper agents use Cursor’s Grok 4.5 High model. It also checks custom agents and blocks explicitly chosen models that are not allowed.

In plain words
What is it for?
Use it in Cursor projects where task launches, subagents, and custom agents must follow one approved model choice.
Why use it?
It keeps model selection consistent across agent work and prevents launches from using an unintended or faster model.

Cursor rule for Cursor

Install

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.

agentmods
npx agentmods add rules/wyattowalsh/agents/cursor-models
Clone the repo
git clone --depth 1 https://github.com/wyattowalsh/agents

Made for: Cursor.

Per session 404 This file is loaded in full into every session.
When invoked 404 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00404 $0.00404
Opus 5 $0.00202 $0.00202
Sonnet 5 $0.00081 $0.00081
Haiku 4.5 $0.00040 $0.00040

Measured 2d ago against content hash 2bd14b506c79, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cursor-models 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 2d 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.

.cursor/rules/cursor-models.mdc · 28 lines

What it actually says

Cursor Model Pin

When calling Task or launching subagents, always set model to cursor-grok-4.5-high.

Do not:

  • Omit model on Task launches (soft rule — Phase B still allows omit so the High parent can inherit)
  • Pass inherit, cursor-grok-4.5-fast, cursor-grok-4.5, composer-*, Claude, GPT, Terra, or Sol slugs
  • Rely on Explore’s faster default — operators SHOULD set CLI exploreSubagentModel: "inherit" in user-owned ~/.cursor/cli-config.json (sync SHALL NOT write cli-config or live state.vscdb)

Custom agents under .cursor/agents/ (and managed home agents) are pinned to cursor-grok-4.5-high.

Layer matrix (SSOT)

Layer Mechanism Omit model Wrong/fast model Hook crash
Soft rule .cursor/rules/cursor-models.mdc Agents must not omit Agents must not pass N/A
Phase A preToolUse Task rewrite Pins to High via updated_input Rewrites + clears alts fail-open (Task proceeds)
Phase B subagentStart allowlist Allow (inherit High parent) Deny if explicit non-High fail-open
Validate/CI validate_hooks + pytest N/A N/A Compensatory: projection must exist

Security Enforce guards (beforeShellExecution, etc.) stay fail-closed (RV-004). Pin hooks stay fail-open.

Changes

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.

  1. 2d ago First seen · 28 lines · 404 tokens per session scan A 2bd14b506c79

Subscribe to this mod's changes

cursor-models is a cursor rule published in the GitHub repository wyattowalsh/agents (5 stars, last pushed 11d ago), licensed MIT. It adds 404 tokens to every session, about $0.0020 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.