agentleFS
InfiniteRoomLabsInfiniteRoomLabs/agent-ops/CLAUDE.md

agent-ops

CLAUDE.md0 starsChanged 5 months agoMITView on GitHub

Infinite Room Labs AI Agency -- a single Claude Code plugin providing 156+ specialized agents with NEXUS multi-agent orchestration. It is the company's open-source toolkit for running the business through AI agent teams. Single plugin: infinite-room-labs@agency. All components live at the repo root.

The file

We split this CLAUDE.md into the rules it gives an agent

8 rules. About 0 are common, 1 is distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    All agents use this extended frontmatter:

    line 69
  2. Very specific:

    Use the new-agent skill, or manually:

    line 110
  3. Very specific:

    Create the file in the correct division directory under agents/

    line 112
  4. Very specific:

    Update registry.yaml at the repo root

    line 114
  5. Very specific:

    If adding a new division, update the agents array in .claude-plugin/plugin.json

    line 115
  6. Very specific:

    Run uv run pytest before committing.

    line 123
  7. Very specific:

    Always use tools/version_bump.py to bump versions: Do not edit the three files by hand -- the script handles plugin/pyproject sync, …

    line 138
  8. Very specific:

    After bumping: edit CHANGELOG.md by hand, add a ## [agency-<new-version>] - YYYY-MM-DD section above the previous one, then git add and …

    line 163

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

What This Is

Infinite Room Labs AI Agency -- a single Claude Code plugin providing 156+ specialized agents with NEXUS multi-agent orchestration. It is the company's open-source toolkit for running the business through AI agent teams.

Repository Structure

Single plugin: infinite-room-labs@agency. All components live at the repo root.

.claude-plugin/plugin.json        # plugin manifest with agent directory declarations
agents/                            # 156+ agent definitions organized by division
  core/                            # CEO, Orchestrator, Reality Checker
  engineering/                     # CTO, VP Eng, DevOps chain, Entropy (secrets) + imported specialists
  design/                          # UI, UX, brand, visual specialists
  marketing/                       # Growth, content, social media, SEO specialists
  sales/                           # Account strategy, pipeline, outbound specialists
  paid-media/                      # PPC, programmatic, social ads specialists
  product/                         # Sprint planning, feedback, behavioral design
  project-management/              # Studio production, project shepherding, experiments
  testing/                         # Evidence collection, benchmarking, API testing
  support/                         # Analytics, finance, infrastructure, legal, support
  research/                        # Competitive intelligence, tech evaluation
  spatial-computing/               # XR, visionOS, Metal, immersive specialists
  game-development/                # Unity, Unreal, Godot, Roblox, narrative design
  specialized/                     # Cross-cutting: compliance, blockchain, data, MCP
skills/                            # 28 user-invocable workflows
  nexus/                           # Full multi-agent orchestration (replaces devops-team)
  nexus-sprint/                    # 2-6 week feature/MVP builds
  nexus-micro/                     # 1-5 day tasks with pre-built runbooks
  load-agent/                      # SUMMON agent persona loading
  end-agent-session/               # SUMMON session teardown
  new-agent/                       # Agent scaffolding
  project-onboard/                 # Repository onboarding
  prompt-refiner/                  # Prompt improvement workflow
  release-prep/                    # Release cycle management
  dep-audit/                       # Dependency auditing
  market-scan/                     # Market research
  web-ui-automation-survey/        # Read-only live-app survey -> automation report
  web-ui-automation-components/    # Survey report -> typed Puppeteer components
  architecture/ code-review/ debug/ deploy-checklist/ documentation/
  incident-response/ standup/ system-design/ tech-debt/ testing-strategy/
                                   # Engineering pack (vendored from Anthropic's
                                   # `engineering` Cowork plugin v1.2.0; links to
                                   # CONNECTORS.md at repo root)
commands/                          # 5 slash commands
  find-agent.md                    # Tag-based agent/skill/command search
  load-agent.md                    # Agent loading command
  end-agent-session.md             # Session cleanup
  publish-package.md               # Package publishing workflow
  trend-watch.md                   # Trend monitoring
hooks/hooks.json                   # SessionStart, PreCompact, PreToolUse hooks
scripts/
  summon.py                        # Agent discovery & session state CLI
  changelog-guard.py               # Commit protection for CHANGELOG
strategy/                          # NEXUS reference material
  nexus-strategy.md                # Full orchestration doctrine
  playbooks/                       # Phase 0-6 playbooks
  coordination/                    # Handoff templates, activation prompts
  runbooks/                        # Scenario-specific execution guides
registry.yaml                     # Master index of all components

Agent Frontmatter Schema

All agents use this extended frontmatter:

---
description: {string}              # Required: trigger text for Claude Code agent selection
model: {opus|sonnet|haiku}         # Required: model tier
tools: {comma-separated list}      # Required: tool access
color: {hex or color name}         # Required: visual indicator
tags:
  function: [{enum list}]          # engineering, operations, finance, research, executive, revenue, creative
  scenario: [{kebab-case list}]    # workflow contexts
  custom: [{kebab-case list}]      # freeform labels
---

Model Tiers

TierUse CaseAgents
opusExecutives, orchestrators, final quality gatesCEO, CTO, VP Eng, Orchestrator, Reality Checker
sonnetMid-level specialists (default for most agents)Engineers, designers, marketers, analysts
haikuSimple tasks, research, validationInterns, basic lookups

NEXUS Orchestration

NEXUS replaces the old /devops-team skill with universal multi-agent orchestration:

ModeSkillAgentsTimelineUse Case
Full/nexusAll12-24 weeksComplete product lifecycle
Sprint/nexus-sprint15-252-6 weeksFeature/MVP builds
Micro/nexus-micro5-101-5 daysSpecific tasks (bug fix, campaign, audit)

Tagging Convention

Standard function values: engineering, operations, finance, research, executive, revenue, creative

Scenario and custom tags: freeform, use kebab-case

Adding Components

Use the new-agent skill, or manually:

  1. Create the file in the correct division directory under agents/
  2. Include YAML frontmatter with description, model, tools, color, and tags
  3. Update registry.yaml at the repo root
  4. If adding a new division, update the agents array in .claude-plugin/plugin.json

Notable Engineering Agents

Entropy (agents/engineering/entropy.md) -- Secrets lifecycle manager. Tracks rotation schedules, reports overdue secrets, and coordinates with bw-sync.sh for Bitwarden-to-Ansible/K8s sync. Invoke when auditing secret age, planning a rotation sprint, or responding to a compromised credential. Tags: engineering, secrets, compliance.

Testing

Run uv run pytest before committing. Testing conventions, the five testing standards (S1-S5), and the not-yet-automated manual checks live in TESTING.md. Standard S1 (every new top-level scripts/*.py ships with its tests/test_<name>.py) is machine-enforced by scripts/test-coverage-guard.py, wired in .claude/settings.json.

Release Workflow

This repo tracks three independent versions:

FileVersionBumped when
.claude-plugin/plugin.jsonplugin versionany change a plugin user would care about
pyproject.toml(synced to plugin.json)every plugin bump -- the version-guard hook will block the commit if these disagree
.claude-plugin/marketplace.json (agency entry)marketplace catalog versionindependent; bump when the marketplace listing itself needs to change

Always use tools/version_bump.py to bump versions. Do not edit the three files by hand -- the script handles plugin/pyproject sync, marketplace independence, and uv lock refresh in one step.

# Most common: same bump across plugin + pyproject (synced pair)
uv run tools/version_bump.py --plugin minor --pyproject minor

# Bump everything together (plugin + pyproject + marketplace entry)
uv run tools/version_bump.py --all minor

# Marketplace entry alone (catalog metadata change, no plugin behavior change)
uv run tools/version_bump.py --marketplace patch

# Pin to an exact version
uv run tools/version_bump.py --plugin set:2.0.0 --pyproject set:2.0.0

# Preview without writing
uv run tools/version_bump.py --all minor --dry-run

Bump-level guidance (Semantic Versioning):

  • major -- breaking change to a plugin component contract (renamed agent, removed skill, hook signature change)
  • minor -- new agent/skill/command/hook/output style, additive behavior
  • patch -- bug fix, doc fix, internal refactor with no consumer-visible change

After bumping: edit CHANGELOG.md by hand, add a ## [agency-<new-version>] - YYYY-MM-DD section above the previous one, then git add and commit. The script does not touch CHANGELOG, the staging area, or git history.

Installation

# Install the agency plugin
/plugin install agency@infinite-room-labs

Cross-Repo Integration

New repos are cut from InfiniteRoomLabs/template-repo, which already carries this file. For an existing repo, .claude/settings.json (committed via the .gitignore allowlist .claude/*, !.claude/.gitignore, !.claude/settings.json) is:

{
  "extraKnownMarketplaces": {
    "claude-plugins-official": {
      "source": { "source": "github", "repo": "anthropics/claude-plugins-official" }
    },
    "infinite-room-labs": {
      "source": { "source": "github", "repo": "InfiniteRoomLabs/agent-ops" }
    }
  },
  "enabledPlugins": {
    "agency@infinite-room-labs": true,
    "superpowers@claude-plugins-official": true,
    "claude-md-management@claude-plugins-official": true
  },
  "enabledMcpjsonServers": ["likec4"]
}

This is the canonical default plugin set. agency is the only plugin this marketplace ships (core@infinite-room-labs never existed); superpowers comes from the official marketplace (obra's superpowers-marketplace is superseded). enabledMcpjsonServers names the project .mcp.json servers a fresh clone may start without a prompt; prefer it over enableAllProjectMcpServers.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.