AgentSkillsCN

draft-life-note

为旅行行程撰写初稿笔记。当您处理 `resources/views/life/trips/*.blade.php` 文件时，这些文件中仅有零星的 `IMG_xxxx.yyy` 行，此技能便会自动激活。

SKILL.md
--- frontmatter
name: draft-life-note
description: >-
  Writes first draft note for a travel trip. Activates when dealing with `resources/views/life/trips/*.blade.php` files which only have lonely `IMG_xxxx.yyy` lines.

Life note writer

When to Apply

Activate this skill when:

  • •Working with resources/view/trips/*.blade.php file
  • •That file doesn't have Russian or English text for every IMG_xxxx image

Context and Style

The author acts as an experienced Urban Observer. The tone is neutral, concise, and informative, occasionally spiced with dry humor or irony. The author is not a typical "tourist" admiring views, but rather an inspector of how the city functions.

Key characteristics:

  • •Genre: Urban exploration / Travel log / Infrastructure review.
  • •Vibe: "How does this work?", "How much does it cost?", "Why is this here?", rather than "Look how beautiful this is."
  • •Evolution: Consistent style since 2007.

Usage Guide

Your goal is to transform a raw list of image filenames into a structured first draft.

1. Context Gathering (Crucial)

Before analyzing any images, you must establish the context of the destination.

  1. •Identify the City: Parse the current filename (e.g., berlin_2016.blade.php -> berlin).
  2. •Find Previous Visits: Search for other files in resources/views/life/trips/ that start with the same city name (e.g., berlin_2017_05.blade.php, berlin_2019.blade.php).
  3. •Read & Digest: Read those files to understand: 3.1. What has already been described? (Do not repeat descriptions of the TV Tower if it was covered in 2016). 3.2. What was the vibe back then? 3.3. Goal: Your new notes should focus on what changed or new details not previously noticed.
  4. •Global Context: Keep in mind your knowledge of other cities. If a trash can looks like one in Tokyo, or a metro gate reminds you of Paris, make that connection.

2. Analyze the Image

For each image (IMG_xxxx.yyy), read it visually and check its metadata (using exiftool). Real images are available in ~/Downloads/buffer/ folder (for example, IMG_1234.jpeg is available as ~/Downloads/buffer/IMG_1234.jpeg).

What to look for (Topics of Interest):

  • •Public Transport: Interiors (seats, screens), ticket machines (interface, price), stations, navigation signage, bus stops.
  • •Road Infrastructure: Signs, markings, traffic lights, surface quality (tiles vs asphalt), curbs, parking meters.
  • •Urban Furniture: Benches, manholes, trash cans/recycling, fences, mailboxes, intercoms.
  • •Retail: Supermarket shelves, prices, receipts, vending machines, distinct packaging.
  • •Accommodation: Interior details (switches, plugs, views from window).
  • •Anomalies: Anything broken, weird, funny, or culturally distinct.

IMPORTANT: You must process images one by one. Do not try to read or analyze multiple images in a single tool call or step, as this may cause the process to hang. Finish one image completely before moving to the next.

3. Determine Description Length (The 3 Tiers)

Tier 1: The Label (Short) Use for: Generic views, simple objects, repeating themes. Length: 1-3 words. Examples:

  • •"Embankment."
  • •"Manhole."
  • •"Evening streets."
  • •"Street signs."

Tier 2: The Observation (Medium) Use for: Things with a specific detail worth noting. Length: 1-2 sentences. Examples:

  • •"Walk-through passage on the green metro line."
  • •"Stop line for cyclists. And a button to trigger the light."
  • •"Access to the central pool is denied without a cap."

Tier 3: The Story / Fun Fact (Long) Use for: Unique objects, funny signs, complex interactions (buying tickets), interactions with people, or cultural shocks. Length: A paragraph or more. Examples:

  • •A story about a funny bus driver who counted passengers.
  • •A detailed explanation of how a confusing ticket machine works or how you got a discount.
  • •A breakdown of a grocery receipt or a specific local product.
  • •Instruction: If the image triggers a specific fact (e.g., "This building looks like X but is actually Y") or a little story, write it down!

4. Output format

<code-snippet name="Example Output" lang="blade"> @ru <p>Tier 3: История о том, как я пытался купить билет в этом автомате. Экран тугой, карты не принимает. Пришлось звать помощь.</p> @endru IMG_1234.jpeg

@ru

<p>Tier 1: Турникеты.</p> @en <p>Turnstiles.</p> @endru IMG_1235.jpeg

@ru

<p>Tier 2: Разметка на полу подсказывает, где стоять в очереди.</p> @en <p>Floor markings suggest where to stand in line.</p> @endru IMG_1236.jpeg </code-snippet>

5. Rules

  • •Minimum Length: Even "simple" descriptions should be at least a word or two. Never leave it empty. Tier 2 descriptions are the most welcome.
  • •Language: Both Russian (@ru) and English (@en) are mandatory.
  • •Cleanup: Delete lines for files that do not exist on disk.
  • •Do not group: Keep one description per image for now.
  • •Image filenames: Keep IMG_xxxx.yyy lines as is.

6. Translation & Localization

  • •Smart Adaptation: Do not translate blindly.
    • •If the image contains English text and the Russian note explains it, do not translate that explanation back to English (it would be meaningless). Instead, provide context, a witty remark, or a different observation in English.
  • •Currency: English text shouldn't contain prices in RUB if the note is not about the country that uses RUB.

Tools

To get metadata:

code
exiftool -json -g -n ~/Downloads/buffer/IMG_xxxx.jpeg