Skip to content

LazyIsEfficient/agentic-os

Repository files navigation

Engineering Heresy — Agentic Framework

A collection of skills and agents for Claude Code that encode engineering workflows, content pipelines, game development, marketing ops, and more into reusable AI playbooks.

Install once, use in any project.

Part of Engineering Heresy by Glenn Eggleton — challenging conventional wisdom in AI and software engineering. Subscribe on Substack →


Install

The one-liners install a pinned release and verify its SHA-256 before extracting anything — see Verifying the download.

  • Current release: v1.4.0
  • Asset: agentic-os-v1.4.0.tar.gz
  • SHA-256: a999d63479e20431c6e30c8079f6d9764d080e0dea6981de43dc74dfbdbe16c9

macOS / Linux

One-liner (no clone required):

curl -fsSL https://raw.githubusercontent.com/LazyIsEfficient/agentic-os/v1.4.0/install.sh | bash

Or from a local clone:

git clone https://github.com/LazyIsEfficient/agentic-os.git
cd agentic-os
./install.sh

Files are copied to ~/.claude/skills/, ~/.claude/agents/, and ~/.claude/commands/. Existing files are not overwritten by default. Add --force to update everything.

Windows (PowerShell)

One-liner (no clone required):

irm https://raw.githubusercontent.com/LazyIsEfficient/agentic-os/v1.4.0/install.ps1 | iex

Or from a local clone:

git clone https://github.com/LazyIsEfficient/agentic-os.git
cd agentic-os
.\install.ps1

Files are copied to %USERPROFILE%\.claude\skills\, %USERPROFILE%\.claude\agents\, and %USERPROFILE%\.claude\commands\. Add -Force to overwrite existing files. (The remote install uses tar, which ships with Windows 10 1803+.)

Verifying the download

The remote install path downloads the pinned release asset and aborts if its SHA-256 does not match the digest embedded in the installer. To verify out-of-band before trusting the one-liner, download the asset and check it yourself:

curl -fsSLO https://github.com/LazyIsEfficient/agentic-os/releases/download/v1.4.0/agentic-os-v1.4.0.tar.gz
# macOS / BSD:
echo "a999d63479e20431c6e30c8079f6d9764d080e0dea6981de43dc74dfbdbe16c9  agentic-os-v1.4.0.tar.gz" | shasum -a 256 -c
# Linux (coreutils):
echo "a999d63479e20431c6e30c8079f6d9764d080e0dea6981de43dc74dfbdbe16c9  agentic-os-v1.4.0.tar.gz" | sha256sum -c

There is intentionally no "track main" remote install path — to install unreleased changes, clone the repo and run ./install.sh from the clone. Maintainers: see RELEASING.md for how the pin is produced.

Custom install path

CLAUDE_DIR=/path/to/.claude ./install.sh
.\install.ps1 -Dest "C:\path\to\.claude"

What gets installed

Directory Contents
~/.claude/skills/ Skill playbooks — invoked with the Skill tool or /skill-name
~/.claude/agents/ Subagent definitions — spawned with the Agent tool
~/.claude/commands/ Slash commands — /skill-new and /agent-new scaffold a new conforming skill or agent; /route recommends the owning skill/agent for a task; /v2-collab runs an in-session multi-agent collaboration pod
~/.claude/workflows/ Workflow definitions backing shipped commands — currently the v2-collab workflow, resolved by /v2-collab
~/.claude/hooks/ PreToolUse hooks (e.g. block-bad-bash.sh)

Ship vs. in-repo-only. The installer copies a curated allowlist, not whole directories. Only the author-facing commands (skill-new, agent-new, route) and the v2-collab command (plus its v2-collab workflow) install into your global namespace. Maintainer-only tooling that lives in this repo — the audit-library / review-gate / plan-clean / triage-findings commands and the other workflows/ (sharded library audit, routing-collision sweep) — is not installed, to avoid polluting your command namespace.


Usage

Invoking a skill

In any Claude Code conversation, reference a skill by name:

Use the test-driven-development skill to write tests for this module.

Or use a slash command if configured:

/test-driven-development

Agents are spawned automatically when Claude Code routes a task (e.g. engineer, code-reviewer), or you can request one explicitly:

Use the security-reviewer agent to audit this PR.

Multi-agent collaboration pod (/v2-collab)

Run a small pod of agents that collaborate over several rounds on one deliverable: a PM frames the work, an engineer builds it, and a reviewer critiques each round until it is approved (or a round cap is hit). Unlike a single agent or a one-shot fan-out, the pod iterates — each round sees the previous round's artifact and the reviewer's notes. It runs entirely in your session as a Workflow (no extra services, no API keys); the produced files are written to a path you name (or ./v2-out/), never into your live library.

Example prompt that triggers it:

/v2-collab Build a single-page marketing site for an AI healthcare startup whose
service ingests your health data, has licensed nurses and doctors analyze it, and
returns personalized recommendations. Write it to pocs/sample.

The roster is configurable; by default it is technical-pm → engineer → code-reviewer, and the last role is the approval gate — the run ends the moment that reviewer approves. Pass a lower round cap for a cheaper first run.

Skills vs Agents

Skills are instruction playbooks — they tell Claude how to do a specific type of work (TDD, debugging, API design). They are stateless and composable.

Agents are role definitions — they give Claude a persona, a tool allowlist, and a mandate (e.g. a full-stack engineer, a security auditor). Agents can invoke skills.

Configure ~/.claude/CLAUDE.md

Skills are available to Claude, but Claude won't automatically reach for them unless instructed to. Add the following to ~/.claude/CLAUDE.md to make Claude check for a relevant skill before responding to any task:

## Skills

You have a library of skills installed at `~/.claude/skills/`. Before responding to any task,
check whether a skill applies and invoke it with the `Skill` tool if so — even if the task
seems simple. Use `using-agent-skills` to discover which skill fits when unsure.

If there is even a 1% chance a skill might apply, invoke it first.

This is the single most impactful configuration step — without it, Claude treats skills as opt-in rather than default.


Skills

Skill Description
adversarial-claims-reviewer Adversarially verify formal/technical claims — math, stats, benchmarks
api-and-interface-design Design stable APIs and interfaces that are hard to misuse
autoresearch Run Karpathy-style autoresearch optimization on any content
blog-post-author Draft a blog post from a filled blog-post-shaper brief
blog-post-shaper Structure a vague blog idea into a well-scoped brief
browser-testing-with-devtools Test in real browsers using Chrome DevTools
ci-cd-and-automation Automate CI/CD pipeline setup and quality gates
cloud-infrastructure Provision or modify cloud resources with infrastructure-as-code
code-review-and-quality Multi-axis code review before merging changes
code-simplification Simplify code for clarity without changing behavior
codebase-cost-estimator Estimate build/dev cost of a codebase by measured LOC and complexity
content-ops Score and evaluate content using an auto-assembled expert panel
content-pipeline Content-production pipeline — quote mining, clip discovery, repurposing, batch gating
context-engineering Optimize agent context setup for quality
conversion-ops AI-powered conversion rate optimization and lead magnet generation
course-author Write lesson content from a filled lesson spec
course-design Turn a course brief into a concrete outline
course-shaper Structure a vague course idea into a well-scoped brief
debugging-and-error-recovery Systematic root-cause debugging workflow
deck-generator Generate professional presentations with AI-generated images
deployment-pipelines Author or review CI/CD pipelines and deployment workflows
deprecation-and-migration Manage deprecation and migration of old systems
devops-engineer Platform and DevOps engineering for Kubernetes, Helm, Pulumi IaC, and CI/CD mechanics
documentation-and-adrs Record decisions and documentation as ADRs
documentation-writer Keep repository documentation accurate and in sync
elevenlabs-tts Convert text to speech using ElevenLabs
finance-ops AI-powered financial analysis and CFO briefings
findings-ledger Record and triage stochastic (Tier 2) review findings for recurrence
frontend-ui-engineering Build production-quality UIs and components
game-balancer Tune game economy curves, progression, and balance
game-concept-creator Generate, evaluate, and refine pitch-quality game concepts
game-design-shaper Structure a vague game idea into a well-scoped brief
game-marketer Market a game via store pages, trailers, and communities
game-monetization-strategist Pick and shape the monetization model for a game
game-systems-designer Design game systems from a locked concept
git-workflow-and-versioning Structure git workflow practices and version control
godot-engineer Build games in Godot 4 with C#
growth-engine Autonomous growth experimentation framework with analysis
iap-manager Design and operate the in-app purchase catalog
idea-refine Refine ideas iteratively through divergent thinking
incremental-implementation Deliver changes incrementally in vertical slices
marketing-shaper Structure a vague marketing request into a scoped brief
meeting-intelligence Extract action items, decisions, and follow-ups from meeting transcripts
outbound-engine Design and optimize cold outbound email campaigns
performance-optimization Optimize application performance with measurement
phaser-engineer Build games in Phaser 3 with TypeScript
planning-and-task-breakdown Break work into ordered, parallel-dispatchable tasks
podcast-ops Podcast-to-everything content pipeline and repurposing
prompt-shaper Structure a vague engineering request into a task brief
release-manager Coordinate release preparation across a monorepo
revenue-intelligence AI-powered revenue intelligence and attribution
rust-engineer Write, review, or architect Rust code
security Scan and redact PII and sensitive data
security-and-hardening Harden code against vulnerabilities and threats
security-engineering Cross-stack security review covering all attack surfaces
seo-ops AI-powered SEO operations and keyword intelligence
shipping-and-launch Prepare production launches with checklists
site-reliability-engineering Operate production systems with SLOs and runbooks
skill-library-review Audit a library of skills and agents
social-growth Write LinkedIn and X promo posts for content
software-design Shape the internal structure of code
source-driven-development Ground every implementation in official documentation
spec-driven-development Create specs before writing code
standards-enforcer Review work against agreed engineering standards
system-architect Design new systems and evaluate architectural trade-offs
team-lead Police work tickets and capture architectural decisions
team-ops AI-powered team performance analysis and intelligence
technical-product-management Make product decisions in a technical context
technical-strategist Set technical direction for an engineering organization
telemetry Opt-in, local-first, privacy-respecting usage telemetry
test-driven-development Drive development with tests first
typescript-analytics Implement analytics with PostHog in TypeScript
typescript-data-engineering Build data pipelines, ETL jobs, and event processors
typescript-quality-engineering Establish cross-cutting test strategy for TypeScript
typescript-testing-backend Write backend tests with Jest and Supertest
typescript-testing-frontend Write frontend tests with Jest and React Testing Library
using-agent-skills Discover and invoke agent skills
ux-design Design or review user interfaces and interactions
ux-research Plan and run user research methods
web3-smart-contract-engineering Write and review Solidity smart contracts
x-longform-post Write long-form X posts in founder voice
yt-competitive-analysis Analyze YouTube channels for outlier videos
yt-shorts-pipeline End-to-end YouTube Shorts production pipeline
yt-shorts-script Generate a YouTube Shorts script from a topic

Agents

Agent Description
adversarial-claims-reviewer Read-only, cold-context adversarial review of formal/technical claims
bigquery-ai-agent Expert data analyst for BigQuery — SQL generation, data interpretation, and insight delivery
blog-post-shaper Blog pipeline — intake, draft, emit asset tasks, and fan out
code-reviewer Read-only multi-axis code review
course-shaper Education pipeline — intake, design, and author lessons
devops-engineer Platform and DevOps engineering for Kubernetes, Helm, Pulumi IaC, and CI/CD mechanics
engineer Full-stack implementation across architecture and shipping
game-design-shaper Game design pipeline — intake through marketing end-to-end
godot-engineer Godot 4 + C# game development
library-reviewer Read-only audit of a skill and agent library
marketer Full-spectrum marketing, content, and sales execution
marketing-shaper Marketing intake — turn a vague goal into a scoped brief
ops-analyst Finance and team operations analyst
phaser-engineer Phaser 3 + TypeScript web game development
prompt-shaper Engineering intake — turn a vague request into a task brief
rust-engineer Principal-level Rust engineering — async services, APIs, workspaces
security-reviewer Read-only cross-stack security audit
technical-pm Product strategy, technical strategy, and engineering leadership
ux-specialist UX design and research
web3-engineer Solidity smart contract development on EVM chains

Commands

Slash commands in .claude/commands/. Only agent-new, route, skill-new, and v2-collab ship to consumers (the last together with its v2-collab workflow); the rest are repo-local maintainer tools.

Command Description
agent-new Scaffold a new conforming agent definition
audit-library Launch the sharded, adversarially-verified skill-library audit
plan-clean Find completed plans in .claude/plans/ and delete them after confirmation
review-gate Run the mandatory build + review pairing gate on the current diff
route Recommend the owning skill/agent for a task
skill-new Scaffold a new conforming skill
triage-findings Tally the findings ledger and propose ratchet targets (human disposes)
v2-collab Run an in-session multi-agent collaboration pod (PM → engineer → reviewer) over rounds on one task

Repository layout

.claude/
├── skills/<skill-name>/
│   ├── SKILL.md          # frontmatter + rules (target <100 lines)
│   ├── references/       # deep-dive docs loaded on demand
│   ├── assets/           # fill-in templates (ADRs, RFCs, briefs)
│   └── scripts/          # runnable helpers
├── agents/<agent-name>.md
├── commands/<command>.md # slash commands (author-facing + maintainer-only)
├── hooks/                # PreToolUse hooks (e.g. block-bad-bash.sh)
├── rules/                # operating doctrine, @-imported by CLAUDE.md
└── workflows/            # multi-agent orchestration scripts

Only a curated allowlist ships to consumers (see What gets installed); CLAUDE.md and rules/ are repo-local and are never installed.


Contributing

Pull requests welcome — see CONTRIBUTING.md for conventions and review gates. The short version: scaffold with /skill-new or /agent-new, run the library-reviewer agent on your diff, and make sure bash scripts/validate.sh passes.


Validating the library

A deterministic, LLM-free validator checks structural invariants — frontmatter completeness, kebab-case names matching their file/dir, no dangling links or @-imports, MEMORY.md length, review-tier wiring (and findings-ledger shape, if present), and that the install scripts ship exactly the curated allowlist.

Enable the pre-commit hook once per clone so the validator runs before every commit:

git config core.hooksPath .githooks

Run it manually any time:

bash scripts/validate.sh

CI enforces the same check on every pull request and push to main, and install.sh runs it before copying anything — a library that fails validation will not install.


License

MIT