prompt-engineering

A set of rules for improving prompts, which are instructions given to an AI model. It diagnoses the draft, applies only useful techniques, and returns a revised prompt with a short list of changes.

In plain words
What is it for?
Use it when you want to optimize or rewrite a prompt for Claude 4.x or another named AI model.
Why use it?
It helps make instructions clearer and more reliable without adding unnecessary complexity or changing their intended variables.

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/phalves23/prompt-engineering-skill/prompt-engineering
Clone the repo
git clone --depth 1 https://github.com/PhAlves23/prompt-engineering-skill

Made for: Cursor.

Per session 37 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,063 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 $0.00037 $0.01063
Opus 5 $0.00018 $0.00531
Sonnet 5 $0.00007 $0.00213
Haiku 4.5 $0.00004 $0.00106

Measured 2d ago against content hash 1a6da38eb0c2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

prompt-engineering 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.

adapters/cursor/.cursor/rules/prompt-engineering.mdc · 76 lines

How it starts

The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Prompt Engineering

Turn a raw draft into a production-grade prompt. Improve the prompt, don't run it — return a rewritten version + a short changelog. Only execute the task the prompt describes if the user explicitly asks.

Operating principle

  • Calibrate effort to complexity. Simple prompt → lean structural rewrite. Complex prompt (reasoning, classification, generation with criteria, agentic) → full structure. Don't over-engineer.
  • If the draft is already good, say so and do the minimum. Don't invent changes to justify a rewrite.
  • Default target model is Claude 4.x. If the user names another (GPT/o-series, Gemini), adjust accordingly (reasoning models do NOT want "think step by step").
  • Preserve every {{variable}} from the original. Convert negative instructions into positive ones. Add motivation to non-obvious instructions.

Workflow

  1. Diagnose — real intent, task type, target model, audience, output format, constraints, dynamic variables, draft weaknesses.
  2. Select techniques — only those that add value (see decision table).
  3. Rewrite in the canonical structure.
  4. Self-review — could a colleague with no context run this without doubt? Fix what fails.
  5. Deliver — optimized prompt + changelog.

Canonical structure (omit sections that don't add value)

  1. Role/persona — one line: who the model is + domain.
  2. Task + objective — what to produce + success criterion (outcome-oriented).
  3. Context + motivation — background + why it matters.
  4. Data/inputs — in XML tags. Long-context (20k+ tokens): long data at the TOP, question at the END (+up to 30% quality).
  5. Instructions — numbered when order matters, positive, explicit scope ("apply to ALL sections, not just the first").
  6. Reasoning<thinking> block for analytical tasks; NOT for reasoning models.
  7. Examples (few-shot) — 3–5 diverse, in <example> tags. Strongest lever for format/tone.
  8. Output contract — exact format (JSON schema, XML, "result only, no preamble"). Say what to do, not what to avoid.

Read the full file on GitHub · 76 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 · 76 lines · 37 tokens per session scan A 1a6da38eb0c2

Subscribe to this mod's changes

prompt-engineering is a cursor rule published in the GitHub repository PhAlves23/prompt-engineering-skill (9 stars, last pushed 2mo ago), licensed MIT. It adds 37 tokens to every session and 1,063 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.