taste-engineer

taste-engineer is an agent for Cursor from ashishpatill/tell-proof. It costs 46 tokens per session (325 once invoked), scanned A, original, MIT.

A specialist prompt for judging visual design findings and turning design directions into typed plans. It can use an optional Gemini refinement, while its parser also works offline for demonstrations and CI.

In plain words
What is it for?
Use it for UI comparison findings, taste verdicts, art-direction presets, and direction-plan parsing in the taste package.
Why use it?
It separates visual judgment from detection and provides a fallback when an external AI service or API key is unavailable. It also validates generated results against known facts.

Agent 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 agents/ashishpatill/tell-proof/taste-engineer
Clone the repo
git clone --depth 1 https://github.com/ashishpatill/tell-proof

Made for: Cursor.

Wrote 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.

agentmods badge for taste-engineer

README.md
[![agentmods](https://agentmods.dev/badge/agents/ashishpatill/tell-proof/taste-engineer.svg)](https://agentmods.dev/agents/ashishpatill/tell-proof/taste-engineer)
Your own site
<a href="https://agentmods.dev/agents/ashishpatill/tell-proof/taste-engineer"><img src="https://agentmods.dev/badge/agents/ashishpatill/tell-proof/taste-engineer.svg" alt="Measured on agentmods" height="20"></a>
Per session 46 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 325 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 5.1 $0.00046 $0.00325
Opus 5 $0.00023 $0.00162
Sonnet 5 $0.00009 $0.00065
Haiku 4.5 $0.00005 $0.00032

Measured 5d ago against content hash fbf8d82ff4fc, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

taste-engineer 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 5d 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/agents/taste-engineer.md · 29 lines

What it actually says

You are Tell's taste engineer. You own judgment — not detection.

Scope

  • packages/taste — TasteVerdict generation, direction plan parsing, mechanical fallback

Rules

  1. Facts from detectors are authoritative — never contradict them in rationale
  2. Reflection loop: draft → validate against facts → refine once → fallback at confidence 0.5
  3. Verdicts: generic | drift | intentional | uncertain
  4. Rationale ≤ 3 sentences, critic voice, no apology
  5. Direction presets: editorial, precision, warm-minimal, bold-contrast, luxury, brutalist, explainer → DirectionPlan + ArtDirection
  6. Parser must work offline for demo (no API key required for CI); Gemini refinement is optional

Prompt contract

System: "You are Tell's taste engine. Classify rendered-UI findings. JSON only."

DoD

  • All findings get a verdict on fixture report
  • Brutalist/intentional section returns intentional with reason
  • Contract test with mocked Gemini JSON passes zod validation

Do not implement UI or MCP — return typed functions for @tell/mcp and web API routes.

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. 5d ago First seen · 29 lines · 46 tokens per session scan A fbf8d82ff4fc

Subscribe to this mod's changes

taste-engineer is an agent published in the GitHub repository ashishpatill/tell-proof (1 stars, last pushed 10d ago), licensed MIT. It adds 46 tokens to every session and 325 once invoked, about $0.0002 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.