bounded-loops

An AI coding agent can tell you it finished when it did not. This runs the agent in a loop that only stops when a separate check — one the agent cannot edit or grade — says the work actually passes, and gives up after a limit you set in advance.

Every run leaves a hash-chained ledger you can re-verify, so "it passed" is a claim you can check rather than take. 69 task templates ship inside the package; 65 need no API key.

Full documentation: github.com/qualixar/bounded-loops

This npm package is a thin launcher for the Python engine.

The engine itself is a Python 3.11+ package. This wrapper lets you run it with one command:

npx bounded-loops new --list
npx bounded-loops new pytest-basic my-loop
npx bounded-loops run my-loop --yes

On first run it detects Python 3.11+, installs the bounded-loops Python package if it isn't already present, and hands off to the real CLI. It does not reimplement the tool in Node — Python 3.11+ must be on your PATH.

Prefer the native install if you already have Python:

pip install bounded-loops
bl new --list
bl new pytest-basic my-loop
bl run my-loop --yes

Full documentation, the 69 loop folders (65 keyless), the nine bounds, and the architecture docs live in the main repository. Clone the repository when you want bl list to browse the full shipped loop catalog.

Apache-2.0.