← Back to Plugins
Tools

Commitment Guard

piiiico By piiiico 👁 48 views ▲ 0 votes

OpenClaw plugin that gates skill and plugin installs on behavioral commitment scores. Content is free to fake — commitment is not.

GitHub

Install

openclaw plugins install clawhub:@amdal/openclaw-commitment-guard

README

# Commitment Guard

OpenClaw plugin that gates skill and plugin installs on behavioral commitment scores.

**Content is free to fake. Commitment is not.**

A maintainer who shipped 847 releases over 12 years is a fundamentally different trust signal than one who published once in 2023. This plugin checks that signal before you install anything.

## What it does

When you run `openclaw plugins install` or `openclaw skills install`, Commitment Guard intercepts the `before_install` hook and:

1. Extracts the npm/PyPI package name from the install event
2. Queries the [Commit](https://getcommit.dev) scoring API for behavioral commitment data
3. Returns findings (info/warn/critical) based on the score
4. **Blocks installs** when packages fall below a configurable threshold or are flagged CRITICAL

## Scoring dimensions

Each package is scored 0-100 on signals that are expensive to manufacture:

| Signal | Weight | What it measures |
|--------|--------|-----------------|
| **Longevity** | 25 pts | How long the package has existed |
| **Download momentum** | 25 pts | Weekly downloads + trend direction |
| **Release consistency** | 20 pts | Version count + recency |
| **Maintainer depth** | 15 pts | Number of active maintainers |
| **GitHub backing** | 15 pts | Linked repository health |

### Risk flags

- **CRITICAL** — sole maintainer + >10M weekly downloads (the exact attack profile from litellm, axios, chalk incidents)
- **HIGH** — new package (<1yr) + high adoption, or sole maintainer + >1M/wk
- **WARN** — no release in 12+ months (abandonment signal)

## Install

```bash
openclaw plugins install clawhub:@amdal/openclaw-commitment-guard
```

Or from npm:

```bash
openclaw plugins install npm:@amdal/openclaw-commitment-guard
```

## Configuration

In your OpenClaw plugin settings:

| Option | Default | Description |
|--------|---------|-------------|
| `apiUrl` | `https://poc-backend.amdal-dev.workers.dev` | Commit scoring API endpoint |
| `blockThreshold` | `30` | Score below this = blocked |
| `warnThreshold` | `50` | Score below this = warning |
| `blockOnCritical` | `true` | Block packages flagged CRITICAL |
| `enabled` | `true` | Toggle on/off |

## Manual audit tool

The plugin also registers a `commitment_audit` tool you can use from any OpenClaw agent:

```
> Audit these packages for supply chain risk: chalk, axios, lodash, express
```

Returns a scored risk table with CRITICAL/HIGH/WARN flags.

## Background

Built after [RFC #55342](https://github.com/openclaw/openclaw/issues/55342) was closed with feedback from @steipete that behavioral reputation belongs as a ClawHub plugin using the `before_install` hook, not in OpenClaw core.

Uses the [Commit protocol](https://getcommit.dev) for behavioral scoring — the same system that flagged chalk, axios, and litellm as CRITICAL before their supply chain incidents.

## License

MIT
tools

Comments

Sign in to leave a comment

Loading comments...