claude

atlas

Role separation with real authority boundaries, deterministic policy enforcement below the model, mission contracts with independently authored acceptance criteria, and a metrics loop.

0 on GitHubby Ranger & Foxv1.0.0MITupdated 2 months agoworkflow
/

Plugin atlas from marketplace atlas-os in RangerAndFox/atlas-os. Commands read from the repo's manifest — 1 month ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (7)

  • atlas-adopt

    Adopt any repository into Atlas governance. Use when setting up Atlas on a new project, or re-running adoption after a toolchain change. Detects the real toolchain from the repo, generates .atlas config, a digest-pinned devcontainer, the validation workflow, CODEOWNERS, and the acceptance-test scaff

  • atlas-criterion-critic

    Adversarially attack each acceptance criterion in a mission draft before promotion — try to build an implementation that satisfies the criterion while leaving the outcome false. Use after the mission linter passes and before asking a human to promote a proposal, and whenever amending a criterion.

  • atlas-design-review

    Render a UI-bearing change in CI, screenshot the declared screens with chromium, and have the design-director review the actual pixels plus the repo's own end-to-end run. Emits a verdict the design gate enforces. Use on user-facing PRs. The director can block; the owner can override.

  • atlas-independent-review

    The independent adversarial review that runs automatically on every PR (via the atlas-review.yml workflow) and gates merge. Produces a plain-language GO / NO-GO a non-technical owner can act on. Use when reviewing a pull request as the required `atlas-independent-review` check, or when running the r

  • atlas-mission-linter

    Check a mission draft against the Definition of Ready before it is promoted. Use whenever authoring or amending a mission, and before asking a human to promote a proposal.

  • atlas-retrospective

    Close out a mission by generating and interpreting its retrospective. Use after a mission is accepted, abandoned, or superseded — every time, without exception.

  • atlas-supervisor

    Run one Atlas mission end-to-end through the roles automatically, stopping only at the two authority gates (promote, merge). Use to drive a mission from a product intent to an opened PR without invoking each role by hand. Session-scoped; holds no authority.

Slash commands (4)

  • /atlas:atlas-accept

    Prepare an acceptance decision packet for the human (does not accept — that is a human CLI action)

  • /atlas:atlas-adopt

    Adopt this repository into Atlas governance (detects toolchain, generates config, devcontainer, CI, gates)

  • /atlas:atlas-mission

    Draft a new mission — Mission Control decides, Mission Author contracts, the critic attacks it, Acceptance Engineer writes the failing tests

  • /atlas:atlas-status

    Show the Atlas state of this project — active mission, unproven criteria, gates, invariants

Subagents (8)

  • atlas-acceptance-engineer

    Converts an approved mission's acceptance criteria into executable tests that fail before implementation. Use immediately after a mission is approved and BEFORE any implementation planning. Must not see the implementation plan.

  • atlas-auditor

    Establishes current reality for a repository or subsystem. Use before planning any mission against unfamiliar or drifted code, and whenever a mission's premise needs verifying. Read-only.

  • atlas-design-director

    Independent design and usability advocate. Reviews whether a change LOOKS right and WORKS as intended, from actually-rendered screenshots of the running UI plus the repo's own end-to-end run. Use on user-facing changes. Can block a user-facing PR; the owner can override. Deliberately blind to the im

  • atlas-engineering-director

    Owns engineering interpretation, architecture, and execution planning for an approved mission. Use to turn a mission contract into a bounded implementation plan, and to judge integration readiness. Plans and reviews; does not implement.

  • atlas-implementer

    Executes one bounded, pre-planned engineering task inside a declared file scope. Use only after the Engineering Director has produced a plan and the Acceptance Engineer has written the tests.

  • atlas-mission-author

    Converts an approved product intent into an implementation-neutral mission contract. Use after Mission Control has decided something should happen but before any engineering planning.

  • atlas-mission-control

    Owns product direction and acceptance. Use to decide what happens next, to classify and scope work, and to judge whether a completed mission actually achieved its outcome. Drafts mission proposals; never implements.

  • atlas-reviewer

    Independent adversarial review of an implementation against a mission. Use after implementation and Director review, when the change touches invariants, security, data, money, or external effects. Deliberately blind to the implementer's account.

Hooks (1)

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/hooks/atlas-guard.mjs"

View source on GitHub

Similar plugins

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

Backend API design, GraphQL architecture, workflow orchestration with Temporal, and test-driven backend development

9 skills · 1 command · 8 agents

wshobson/agents 42,461

beads

claudecodex

AI-supervised issue tracker for coding workflows

1 skill · 1 agent · hooks

gastownhall/beads 28,698

MiniMax AI skills library for frontend, fullstack, Android, iOS, shader, GIF sticker, document, presentation, spreadsheet, and multimodal media workflows

18 skills

MiniMax-AI/skills 13,828

worktrunk

claudecodex

Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows. This plugin provides configuration guidance (LLM commit messages, project hooks, worktree paths), automatic activity tracking (🤖/💬 indicators in `wt list` showing active Claude sessions), and a `/wt-diff` pane showing each worktree's changes.

2 skills · hooks

max-sixty/worktrunk 12,481

gsd-core

claude

GSD Core is a meta-prompting, context engineering, and spec-driven development system for AI coding agents.

72 skills · 72 commands · 64 agents · hooks

open-gsd/gsd-core 11,494

Maintain atlas?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex pluginslisted