Skip to content

AI AGENTS

VD $ VD171 @ Priv8 edited this page Nov 18, 2026 · 1 revision

Languages: English | العربية | Čeština | Deutsch | Español | فارسی | Français | हिन्दी | Bahasa Indonesia | Italiano | 日本語 | 한국어 | Nederlands | Polski | Português | Русский | Svenska | ไทย | Türkçe | Українська | Tiếng Việt | 中文


VD Infos for AI agents

VD Infos is built to be read by machines, not only by people. A device method debugger produces a lot of structured evidence - every value, read through every method, lined up for comparison - and that evidence is far more useful to an agent as data than as a screenshot. So the whole result is available as one machine-readable document, and the app is designed around that from the ground up.

If you are an AI coding assistant, an LLM tool, an analysis script or any automation that reaches the device through a shell, this page is for you.

Why it is agent-friendly

  • One document, not a screen. The entire scan - every probe, every lens, every verdict - is a single JSON object. No OCR, no screen scraping, no tapping through collapsed cards.
  • The same data the UI has. The file is not a summary. It is exactly what the app holds internally: the screen shows the first reading of a collapsed item, the file has all of them.
  • Deterministic shape. The structure is stable and documented (see EXPORT-FORMAT), so a parser written once keeps working.
  • No network, no account, no state to manage. The app reads on-device and writes a file. Your agent reads the file. That is the whole contract.
  • Offline by design. It needs no internet, so it runs the same inside a locked-down or air-gapped environment an agent might be operating in.

The one-command read

Every scan writes its full report to the app's private storage. After a scan has run, an agent reads the complete result with a single command:

su -c cat /data/data/ru.vd171.vdinfos/files/last_snapshot.json
  • What it contains: every probe with its verdict and the value from every lens, in the same JSON as the export button.
  • When it updates: at the end of every scan. Run a scan first; the file always holds the latest one.
  • Requirement: root, because the file lives in the app's private storage. Without root, use the in-app share / save as JSON action instead and read the file the agent is handed.

A typical agent flow

  1. Make sure the app is installed (the SDK_35 APK from Releases).
  2. Trigger a scan - a person taps scan, or an automation launches the app's scan screen - and wait for it to finish.
  3. Read files/last_snapshot.json with the command above.
  4. Parse it with the export format as your schema: iterate the probes, read each one's verdict, and line up the per-lens values.
  5. Act on the structured result in your own workflow.

The agent decides what the readings mean for the task at hand; VD Infos supplies the evidence, cleanly and completely.

What the agent gets, and what it does not

  • The report carries every reading and every verdict, so an agent has the full picture the app built.
  • Identity-bearing values (serials, IMEI, MAC addresses, advertising IDs and the like) are masked in the shared/exported report, so a report handed around stays free of personal identifiers while keeping its verdicts and structure. The on-device last_snapshot.json, readable only by the app or root, keeps the real values for a local agent that needs them.

Documentation, written to be parsed

This wiki is structured for machine reading as much as for people: short sections, stable headings, explicit field names, and a dedicated export format page an agent can treat as a schema. If you are building a tool on top of VD Infos, start there.

Related pages


This documentation was generated with Claude (Anthropic) and is maintained by VD171. Corrections and improvements are welcome; see CONTRIBUTING.md.

Clone this wiki locally