sabahattink/full-stack-hq
v1.4.0
Permission-first, tool-agnostic engineering guidance for Claude Code, Google Antigravity IDE, and OpenAI Codex.
Changelog
All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[Unreleased]
[1.4.0] - 2026-10-05
Added — Project installs for every agent on a team
--project DIR/-Project DIRinstalls the shared rules into a repository:AGENTS.md(read by Codex, Cursor, and GitHub Copilot) plus one-line imports inCLAUDE.md(Claude Code) andGEMINI.md(Gemini CLI)- The installer manages only a marked block, keeps existing content and line endings, and updates the block in place when re-run
--uninstall/-Uninstallremoves the blocks and restores the original files; files that held only the block are deletedadapters/project/rules.md, rendered identically by both build scriptsbootstrap.ps1accepts-Projectand-Uninstall- Smoke tests cover project install, idempotent re-runs, CRLF preservation, and byte-for-byte uninstall on Windows and Ubuntu
- Validators keep the project
AGENTS.mdwithin Codex's 32 KiB budget
Added — Documentation
docs/COMPARISON.mdcomparing Full Stack HQ with rule-sync tools and skill frameworks, including when another tool is the better fitREADME.tr.mdTurkish summary
Fixed
install.ps1no longer fails on Linux or macOS whenUSERPROFILEis unset
Changed
- CI uses
actions/checkout@v7andactions/setup-node@v7; Dependabot keeps GitHub Actions current
Added — Discoverability
llms.txtwith a plain-text project summary and doc index for AI assistantsdocs/FAQ.mdanswering the most common questions, such as how to stop an agent from editing files without approval and how to share one set of rules across Claude Code, Codex, and AntigravityCITATION.cffso GitHub offers citation metadata- README introduction that states what the project is, who it is for, and its repository and plugin names
- Validators check that every repository link in
llms.txtresolves
[1.3.0] - 2026-10-04
Added — One-command installs
- Claude Code plugin marketplace (
.claude-plugin/): install with/plugin marketplace add sabahattink/antigravity-fullstack-hqand/plugin install full-stack-hq@full-stack-hq, without cloning - Session-start hook that loads the shared rules plus a small plugin adapter
(
adapters/claude/plugin.md); workflows are exposed as namespaced commands bootstrap.shandbootstrap.ps1one-line installers that fetch a temporary shallow checkout, pass every option to the installer, and clean up- Validator checks for the Claude manifests, version parity with
plugin.json, the hook sources, and the 10,000-character hook budget - Smoke tests now install through the bootstrap scripts
- CI validates the Claude plugin manifests with
claude plugin validate --strict
Fixed
install.shno longer exits when a replace prompt reads end-of-input; it keeps the existing file instead
[1.2.0] - 2026-09-20
Added — Launch and feedback kit
- PNG derivatives of the repository visuals for Dev.to and external image proxies
docs/LAUNCH_PLAYBOOK.mdwith channel-specific copy, rollout sequencing, and measurement guidancedocs/DEMO_SCRIPT.mdfor a real 45-second validation and dry-run walkthrough- Installation-feedback issue template for host-specific reports
- Dev.to cover metadata and PNG-based article visuals
- Portable
plugin.jsonpackage boundary for the existing canonical skills - Portable package guide and current Codex instruction-discovery notes
- Read-only
scripts/doctor.*diagnostics for host commands, target paths, and Codex override precedence - Isolated all-host installer smoke tests in
scripts/smoke-test.* - First-run recipes covering dry-runs, isolated tests, read-only diagnostics, and installation feedback
Improved — First-run discovery
- Added a safe 60-second README preview that does not touch the real home directory
- Added direct README links to the installation feedback form and community discussions
Changed — Repository experience
- Added manual workflow dispatch and concurrency control to validation CI
- Added direct documentation links for OpenAI Agent Plugins and the Agent Skills specification
- CI now runs the installer smoke test on Windows and Ubuntu after validation
[1.1.0] - 2026-09-16
Added — Tool-agnostic core and Codex support
core/rules/common.mdas the shared policy source of truth- Thin host adapters for Google Antigravity IDE, Claude Code, and OpenAI Codex
- Codex
AGENTS.mdrendering and TOML custom-agent adapters for all 10 agents - Workflow-to-skill bridge so all 10 workflows are reusable on every host
- Repository-level
AGENTS.mdwith contribution and verification guidance
Updated — Installers and migration safety
- Added
--only-codex/-OnlyCodex - Added
--dry-run/-DryRun,--backup/-Backup, and--check/-Check - Added
--target-root/-TargetRootfor isolated installer smoke tests - Switched Antigravity global agents and skills to documented
~/.gemini/config/paths while refreshing an existing legacy tree as a compatibility bridge - Existing managed agent and skill files are kept unless Force is selected
- Warns when Codex
AGENTS.override.mdwould take precedence over the installed global rules - Added source and adapter validators for PowerShell and Bash
Added — Continuous validation
- Added a least-privilege GitHub Actions job for Windows and Ubuntu validation
- Normalized workflow metadata from mixed
command/triggertocommand - Removed accidental host branding from portable examples
- Added Antigravity rule-size and workflow/skill collision guards
- Kept PowerShell and Bash renderer output equivalent
[1.0.0] - 2026-05-15
Added — Claude Code Support
claude/CLAUDE.md— 12-section permission-first global rules for Claude Code- Permission-first workflow with explicit approval keywords
- 10 agent role definitions
- Full tech stack defaults (Next.js 15, NestJS, Prisma 6+, TypeScript 5+)
- Code style enforcement (no semicolons, single quotes, 2 spaces)
- Git conventions (Conventional Commits)
- Testing philosophy (Vitest, Jest, Playwright)
- Security checklist
- Claude Code workflow commands
Added — Skills (24 new, total 28)
react-best-practices— Component patterns, hooks, memoization, error boundariestypescript-patterns— Generics, utility types, discriminated unions, type guardstailwind-patterns— Tailwind CSS v4, button variants, dark mode, design tokensfrontend-design— Component design principles, visual hierarchy, spacing systemsweb-design-guidelines— WCAG 2.1 accessibility, semantic HTML, Core Web Vitalsbackend-dev-guidelines— NestJS layering, error handling, structured loggingsoftware-architecture— Clean Architecture, SOLID, domain modeling, CQRSapi-design-patterns— REST conventions, response envelopes, pagination, OpenAPItest-driven-development— Red-Green-Refactor, Vitest setup, mock factoriessystematic-debugging— Scientific method, stack traces, production playbookwebapp-testing— Playwright POM, React Testing Library, MSW mockingdocker-patterns— Multi-stage Dockerfiles, docker-compose dev/prodgithub-actions— CI pipeline, E2E testing, deploy workflowsdeployment-guide— Vercel + Railway/Fly.io, health checks, zero-downtimeauth-patterns— JWT access/refresh tokens, Passport.js, httpOnly cookiessecurity-checklist— OWASP Top 10, rate limiting, audit logging, CORSdocx-official— Word document generation with docx librarypdf-official— PDF generation with PDFKit and Puppeteerpptx-official— PowerPoint generation with pptxgenjsxlsx-official— Excel generation with ExcelJSbrainstorming— Six Hats, RICE prioritization, ADR template, trade-off matrixskill-creator— How to create new skills — format, frontmatter, quality checklistcode-review-patterns— P0-P3 severity, comment writing, PR review flowgit-workflow— GitHub Flow, conventional commits, conflict resolution
Added — Workflows (4 new, total 10)
/status— Progress checkpoint, current state, blockers, next steps/preview— Pre-commit review, quality check, security scan, commit draft/orchestrate— Multi-agent coordination, parallel streams, handoff tracking/ui-ux-pro-max— Visual audit, 10 style directions, accessibility, interactions
Updated — Install Scripts
install.sh— Added--only-claudeand--only-antigravityflagsinstall.ps1— Added-OnlyClaudeand-OnlyAntigravityswitches- Auto-detects installed IDEs and configures accordingly
Updated — README
- Professional redesign with dual IDE presentation
- Collapsible install sections (Mac/Linux + Windows)
- Real conversation example demonstrating permission-first workflow
- Agents table with trigger phrases
- IDE feature comparison table
[0.1.0] - 2026-01-20
Added — Initial Release
gemini/GEMINI.md— Permission-first global rules for Antigravity IDE- 10 specialist agents: frontend, backend, database, architect, code-reviewer, test-engineer, security-auditor, devops-engineer, performance-optimizer, documentation-writer
- 4 skills: nestjs-patterns, prisma-workflow, nextjs-app-router, prompt-engineering
- 6 workflows: /brainstorm, /plan, /debug, /create, /enhance, /test
- Install scripts for Mac/Linux and Windows
- Docs: Setup guide, Customization guide, Contributing guide