SMOKE-TEST

A check for confirming that a project's Cursor subagents use their intended AI models rather than the parent chat model or incompatible model labels. It checks the configured agent files and expected model mapping.

In plain words
What is it for?
Use it to inspect `.cursor/agents/` model settings and verify the configured subagent types before relying on the prd-pr workflow.
Why use it?
It helps reveal silent model substitution, where Cursor uses a different model even though a task appears to succeed. This matters when the workflow depends on particular models for exploration, planning, implementation, or review.

Agent

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 agents/wei6bin/skills/smoke-test
Clone the repo
git clone --depth 1 https://github.com/wei6bin/skills
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,671 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00000 $0.01671
Opus 5 $0.00000 $0.00835
Sonnet 5 $0.00000 $0.00334
Haiku 4.5 $0.00000 $0.00167

Measured 2d ago against content hash 70b1c4531e11, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

SMOKE-TEST 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.

prd-pr-cursor/agents/SMOKE-TEST.md · 158 lines

How it starts

The opening of the file, as written. The whole thing — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.

prd-pr Cursor — model smoke test

Verify that project subagents (.cursor/agents/) run on the intended models, not the parent chat model and not the plugin’s Claude Code sonnet/opus/haiku shorthands.

Expected mapping (this repo)

subagent_type Frontmatter model: UI name (approx.)
code-explorer composer-2.5 Composer 2.5
code-architect claude-opus-4-8 Claude 4.8 Opus
plan-reviewer gpt-5.5 GPT-5.5
impl-backend composer-2.5 Composer 2.5
impl-frontend composer-2.5 Composer 2.5
impl-simplify composer-2.5-fast Composer 2.5 Fast (cheap pass)
test-plan-walker inherit Same as parent

Prerequisite: In Cursor Settings → Models, enable every model above. Opus 4.7 and GPT-5.5 may require Max Mode on your plan. If a model is blocked, Cursor substitutes another model silently — the smoke test will look “green” but wrong.

Confirm slugs match your picker (copy ID from Settings). If a slug fails, try the variant shown in the UI (e.g. gpt-5.5-fast, claude-opus-4-7-fast).


Step 0 — Static check (no LLM)

From repo root:

grep -H '^model:' .cursor/agents/*.md | grep -v SMOKE

Expect the table above. Plugin cache must not be your only agent source:

ls .cursor/agents/code-explorer.md

Step 1 — Parent chat setup

  1. Open n-lite in Cursor.
  2. Start a new Agent chat.
  3. Pick a parent model that is different from subagents (e.g. Composer 2.5 Fast or anything except Opus-only), so inherit on test-plan-walker is distinguishable.
  4. Ensure prd-pr plugin is enabled; do not use agent_type: "prd-pr:…" in prompts.

Step 2 — Paste smoke prompt (parent)

Read the full file on GitHub · 158 lines

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 · 158 lines · 0 tokens per session scan A 70b1c4531e11

Subscribe to this mod's changes

SMOKE-TEST is an agent published in the GitHub repository wei6bin/skills (2 stars, last pushed 10d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,671 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.