AGENT SKILL

prompt-engineering-guidance

Team-level guidance for prompt engineering — standards, review processes, prompt libraries, versioning, and rollout practices. Use when standardizing how an organization writes and maintains prompts.

License 许可
MIT
所属仓库
TanNhatCMS/jarvis
最近更新
2026-11-07
Stars
★ 0
Forks
1

01 前言 Front Matter

name
prompt-engineering-guidance
category
ai-research
description
Team-level guidance for prompt engineering — standards, review processes, prompt libraries, versioning, and rollout practices. Use when standardizing how an organization writes and maintains prompts.

03 SKILL.md 正文

Prompt Engineering Guidance (Team Practice)

Individual prompt skill doesn't scale; teams need shared standards. This skill covers the organizational side: conventions, review, libraries, versioning, and rollout — so prompts become maintained assets instead of tribal knowledge.

Overview

Treat prompts like code: they live in version control, go through review, have tests, and follow style conventions. A prompt library gives teams reusable, vetted components (classifiers, extractors, summarizers) instead of everyone reinventing them. Rollout practices — staging, canary, rollback — apply because prompt changes are behavior changes. The goal is a team where anyone can find, understand, and safely modify any prompt.

When to use

  • Multiple people or teams writing prompts for production systems.
  • Prompts scattered across codebases, notebooks, and chat logs with no ownership.
  • Inconsistent quality: some prompts tested, others written once and forgotten.
  • Preparing to scale LLM features: you need process before volume.

Core concepts

  • Prompt standards: conventions for structure (role/task/format sections), naming, and documentation headers. Consistency makes prompts reviewable.
  • Review process: prompt changes reviewed like code — for clarity, test coverage, cost impact, and safety. A second pair of eyes catches what the author can't.
  • Prompt library: versioned, tested, reusable prompts for common jobs. Each entry has a purpose statement, test set, and owner.
  • Versioning and changelog: prompts versioned alongside the code that calls them; changes logged with rationale and measured impact.
  • Staged rollout: new prompts go to shadow/staging first, then a canary slice of traffic, then full rollout — with rollback ready.
  • Ownership: every production prompt has an owner responsible for its tests, performance, and periodic re-validation.

Practical workflow

  1. Audit current state: inventory all production prompts, their owners (or lack thereof), and test coverage.
  2. Establish the standard: structure template, documentation header, and review checklist. Keep it to one page.
  3. Migrate the highest-traffic prompts first: add tests, assign owners, version them.
  4. Build the library incrementally: extract reusable prompts as they're proven, with their test sets.
  5. Institute staged rollouts for prompt changes: shadow → canary → full, with success metrics defined up front.
  6. Schedule re-validation: prompts re-tested on model updates and quarterly; owners accountable.
Prompt header template:
# <name> v<version> — owner: <person>
# Purpose: <one line>
# Inputs: <variables>   Outputs: <format/schema>
# Tests: <link to test set>   Last validated: <date> on <model>
# Changelog: <version>: <what changed, why>

Common pitfalls

  • Process without adoption: a beautiful standard nobody follows. Start with the highest-impact prompts and prove value.
  • Over-standardization: rigid templates that don't fit creative tasks. Standards for structure and testing, freedom for content.
  • No ownership: "everyone owns it" means nobody maintains it. Name an owner per prompt.
  • Skipping staging: pushing prompt changes straight to production. Prompt changes are behavior changes — roll them out like one.
  • Library rot: a prompt library nobody updates becomes a museum. Tie library entries to re-validation schedules.
  • Review theater: rubber-stamping. Reviewers need the test set and before/after outputs, not just the prompt text.

04 同仓库更多 Skills

TanNhatCMS
TanNhatCMS/jarvis

web-scrape

Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open/fetch fails on JS-heavy pages, or when you need clean markdown/structured extraction from a site. Requires FIRECRAWL_API_KEY.

★ 0
TanNhatCMS
TanNhatCMS/jarvis

design-taste-frontend

tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML/CSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, "AI-made", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.

★ 0
TanNhatCMS
TanNhatCMS/jarvis

test-driven-development

Use when implementing any feature or bugfix, before writing implementation code

★ 0
TanNhatCMS
TanNhatCMS/jarvis

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

★ 0
TanNhatCMS
TanNhatCMS/jarvis

requesting-code-review

Use when completing tasks, implementing major features, or before merging to verify work meets requirements

★ 0
TanNhatCMS
TanNhatCMS/jarvis

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).

★ 0