Kibertum

Tausik Core — AI skill for Claude Code

AI community

AI coding agents can't quietly fake "done".

How to install Tausik Core

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Kibertum/tausik-core and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Tausik Core does

AI coding agents can't quietly fake "done". Signed ed25519 receipts prove the gates actually ran — bound to the commit, verifiable offline. Fail-closed hooks block the edit, not the PR. Harness engineering over Claude Code, Cursor, Codex, Qwen, OpenCode.

Alternatives in AI

Hol Guard — Open-source antivirus for AI agents: block risky tools, secret access, prompt injection, malicious packages, M 485 ★
Prismor — Self-hosted runtime control plane for AI agents 334 ★
Codeflow — OpenVibeCoding is the command tower for AI engineering: plan, delegate, track, resume, and prove long-running 206 ★

README

**English** | [Русский](README.ru.md)

TAUSIK

**AI agents that can't *quietly* fake "done."**

TAUSIK is a discipline layer for AI coding agents. It turns the agent's word — "tests pass," "the task is done" — into something you can actually verify. Plan before code, ship with proof, remember every decision. Not suggestions the agent can ignore: a discipline rail that refuses the easy shortcut and makes the ones it can't refuse visible and recorded — with tamper-evidence against outside edits, not a firewall that claims to stop a determined agent.

[![v1.8.0](https://img.shields.io/badge/version-v1.8.0-blue.svg)](https://github.com/Kibertum/tausik-core/releases) [![signed receipts: ed25519](https://img.shields.io/badge/signed%20receipts-ed25519-6f42c1.svg)](docs/en/receipts.md) [![7115 tests](https://img.shields.io/badge/tests-7115-brightgreen.svg)](#proof-tausik-built-tausik) [![coverage 76%](https://img.shields.io/badge/coverage-76%25-green.svg)](#proof-tausik-built-tausik) [![0 dependencies](https://img.shields.io/badge/dependencies-0-brightgreen.svg)](#whats-inside) [![License: Apache 2.0](https://img.shields.io/badge/license-Apache%202.0-blue.svg)](LICENSE) [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-3776AB.svg)](https://python.org)

Without TAUSIK / With TAUSIK

Your agent does this	TAUSIK does this
Says "I'll just refactor this" and edits 30 files	No active task → BLOCKED. No code edits until a task is open.
Declares "done" with nothing to show for it	QG-2 blocks the close. Every acceptance criterion needs evidence.
Reports a green build you have to take on faith	ed25519 signed receipt. The green is cryptographically bound to the gate and the commit — an outside edit to the database can't forge or replay it.
Tries the same broken approach for the third time	Project memory. Failed approaches are recorded; the agent sees what didn't work.
Quietly skips the test/lint pipeline