---
title: "colossus"
description: "First-party Colossus scheduling, development, plugin authoring, and evidence-based security review Agent Skills."
canonical: https://agentpluginsdirectory.com/plugins/colossus
last-updated: 2026-10-21
---

# colossus
First-party Colossus scheduling, development, plugin authoring, and evidence-based security review Agent Skills.
- Slug: colossus
- Publisher: Colossus
- Repository: https://github.com/obscuritylabs/Colossus
- Manifest: bundled-plugins/colossus/plugin.json
- Version: 0.11.9
- License: Apache-2.0
- Category (editorial): security
- Skills: 6 (browser, coding, computer-use, plugin-authoring, schedule-task, security-review)
- MCP servers: 0
- Stars: 5
- Repository created: 2026-05-14
- Repository last pushed: 2026-10-21
- Publisher type: Organization
- Listing: https://agentpluginsdirectory.com/plugins/colossus
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What colossus does, in the publisher's words

AI agent operations for real work in mission environments.

Colossus is an alpha-stage runtime for organizations that need agents to act under explicit authority, inside enforceable boundaries, with durable evidence of what happened. It brings models, tools, policy, approvals, sandboxing, state, and audit into one system that can operate online or offline.

The project is being built for serious enterprise and public-sector deployment, including regulated, disconnected, and security-sensitive environments. It ships as a Rust binary with a CLI and terminal UI; a Tauri desktop app and authenticated SDKs use the same runtime and security model.

- Accountable effects. Requested actions, decisions, approvals, execution, release, and uncertain outcomes produce durable evidence.
- Bounded execution. Strict tool schemas, one-use permits, sandbox profiles, resource ceilings, and quarantined output keep authority explicit.

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/obscuritylabs/Colossus/HEAD/README.md

## Skills

- browser
- coding: Implement and verify scoped software changes with repository evidence.
- computer-use
- plugin-authoring: Create or update portable Agent Plugins with Agent Skills or MCP servers, validate them with Colossus, and package them as OCI artifacts for registry or offline distribution.
- schedule-task
- security-review: Review trust boundaries, authorization, containment, secrets, and recovery behavior.

Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.