Analysis updated 2026-05-18
Stress test a decision or plan by having an AI challenge your own stated reasons for it.
Uncover the real motivation behind a choice you have been rationalizing to yourself.
Practice being questioned using techniques drawn from documented interrogation doctrine, applied to yourself.
Get unstuck on a project or decision that has been stalled by an unexamined excuse.
| baranskyi/interrogation-partner | aaklon/akinator | abhinavshrivastava950/montara | |
|---|---|---|---|
| Stars | 9 | 9 | 9 |
| Language | — | Go | Python |
| Setup difficulty | easy | moderate | moderate |
| Complexity | 1/5 | 3/5 | 4/5 |
| Audience | general | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Installs as a Claude Code plugin with a single marketplace command.
Interrogation Partner is a Claude Code plugin, packaged as an Agent Skill, that questions you about your own decisions, claims, or plans until the polished story you tell yourself gives way to what is actually going on. Rather than accepting your first explanation and helping you act on it, it treats that explanation as a possible cover story and looks for the gap between what you said and what is actually true. A session moves through six phases. It starts by stating that you can stop it at any time with one word, then lets you tell your story uninterrupted. From there it converts vague statements into specific, checkable facts, looks for evasions or places where your story gets suspiciously smooth, and confronts you with two of your own statements that cannot both be true at once. The goal is a single honest sentence you had been avoiding saying, after which the pressing stops. Its only tool is using your own words against you: it never gives advice, plans, or tells you what you should do. The questioning techniques are explicitly drawn from documented interrogation methods, including approaches associated with the CIA, wartime interrogator Hanns Scharff, the UK's PEACE model, and the cognitive interview technique, reframed as ways to challenge your own reasoning rather than someone else's. You can even ask to be questioned in the style of a specific one of these methods. The README is direct that coercive historical techniques are documented only as cautionary examples of what fails, not as methods this skill actually uses. The project sets firm ethical boundaries: you can only ever interrogate yourself, with your own consent, and the skill refuses outright if asked to help interrogate another person, such as an employee or a suspect. There is no coercion, only pressure built from unanswerable contradictions in what you already said, and you can stop the session at any time. It installs as a Claude Code plugin with a single marketplace command, or manually by copying the skill folder into your Claude Code or Claude Desktop skills directory. It works in both English and Russian and is released under the MIT License.
A Claude Code skill that questions you about your own decisions until your polished story gives way to what is actually true.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.