Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.appliedaifoundation.org/llms.txt

Use this file to discover all available pages before exploring further.

​
What it does

The Image Analysis skill runs AI vision over any image — a single photo, a folder, or a vessel inspection set — and produces a structured assessment. It auto-detects whether the image is maritime (equipment, machinery, vessel surfaces) and switches between two modes:
  • General mode — describes subject, objects, context, and any visible text
  • Vessel mode — assesses equipment condition, detects defects, classifies severity, renames photos to descriptive filenames, and writes a searchable index

​
When to use it

  • “Analyze this image”
  • “What is this?”
  • “Describe photo”
  • “Process vessel photos”
  • “Analyze the inspection batch”
  • “What equipment is in this photo?”

​
Modes

​
General mode

For ad-hoc images. Returns a plain-language description: subject, key objects, environment, any readable text, useful context. Good for screenshots, document captures, quick “what am I looking at?” checks.

​
Vessel mode

For inspection photo sets. The skill:
  1. Classifies the equipment shown (engine room, deck, accommodation, cargo system, etc.)
  2. Rates equipment condition on a 5-point scale: good / fair / poor / damaged / critical
  3. Identifies visible defects with location and likely cause
  4. Suggests a descriptive filename so photos become self-explanatory in the file system
  5. Writes a structured index entry covering equipment, condition, defects, and metadata
The result is a fully audited inspection set — a TSI can scan the index, sort by condition, and jump straight to anything flagged poor or worse.

​
How it works

The skill calls a unified vision engine (Gemini or OpenRouter, depending on configuration) with mode-specific prompts and configuration. Maritime detection runs first; if maritime content is present, vessel mode kicks in automatically. For large batches, the user is asked first which file action to take — rename in place, create renamed copies, or analyze without changing files.

​
What it produces

ModeOutput
GeneralPlain-language description per image
VesselPer-image equipment classification, condition rating, defect list, suggested filename, JSONL index of the whole batch

​
Why it’s split into two modes

Generic image-description prompts produce vague results on engineering photos; condition-rating prompts produce nonsense on screenshots. Detecting the domain first, then routing to the right prompt, gets useful output for both cases without forcing the user to specify mode.
  • pdf-vision-extractor — paired skill for diagrams and tables inside PDFs
  • image-generator — counterpart for generating images instead of analysing them