From omni-sw
Project Setup for Omni-SW Agency — Configures a project directory with all base files and hooks required by the SDLC Agency plugin before the orchestrator can run. Creates CLAUDE.md, agent_docs/ directories, notification hooks, and validates the environment. Use when: (1) setting up a new project for the agency (/init-omni-sw), (2) user says "setup", "configure project", "init project", "prepare project for agency", "install hooks", "configure hooks", "setup omni-sw", "configure plugin", or (3) before running the orchestrator for the first time on a project. This skill should also trigger when the user wants to verify or fix the agency configuration in an existing project.
How this skill is triggered — by the user, by Claude, or both
Slash command
/omni-sw:init-omni-swThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Configure a project directory for the SDLC Agency. This runs in the main session (not as a subagent) so that `AskUserQuestion` works correctly.
Configure a project directory for the SDLC Agency. This runs in the main session (not as a subagent) so that AskUserQuestion works correctly.
Find this skill's own SKILL.md to derive the CLI path:
Glob pattern: "**/init-omni-sw/SKILL.md"
From the SKILL.md path, the CLI is at ../../shared/scripts/agency_cli.py relative to the init-omni-sw/ directory.
Verify it exists with ls, then store the resolved path as {CLI}.
If the user provided a path as argument, use that as {PROJECT_ROOT}. Otherwise, use the current working directory.
Check the directory contents:
ls -la "{PROJECT_ROOT}" 2>&1
.git)Use AskUserQuestion to collect project details. Ask all in a single call:
Then run the setup command with the answers:
MSBUILD_EXE_PATH="" python "{CLI}" setup --scan-root "{PROJECT_ROOT}" --name "{name}" --stack "{stack}" --type "{type}" [--no-tests]
The project has code but no CLAUDE.md. Use the /repo-map skill to scan the codebase first:
/repo-map {PROJECT_ROOT}
The repo-map will discover the stack, structure, and generate CLAUDE.md. After it completes, run setup to create directories and hooks:
MSBUILD_EXE_PATH="" python "{CLI}" setup --scan-root "{PROJECT_ROOT}"
The project already has a CLAUDE.md. Just run setup to ensure directories and hooks are in place:
MSBUILD_EXE_PATH="" python "{CLI}" setup --scan-root "{PROJECT_ROOT}"
After setup completes, run verify to confirm everything is ready:
MSBUILD_EXE_PATH="" python "{CLI}" setup --scan-root "{PROJECT_ROOT}" --verify
Report the results to the user. Pay attention to the hooks status:
If hooks were just created ("action": "created" or "action": "merged"), tell the user:
Hooks were installed. To activate them, either:
- Run
/hooksin Claude Code and approve the notification hook, OR- Start a new Claude Code session (hooks load at startup)
Then run
/orchestrator "<objective>"to start the SDLC.
If hooks were already installed ("action": "already_installed"):
Everything is ready. Run
/orchestrator "<objective>"to start.
npx claudepluginhub vitorandtxr/omni-sw --plugin omni-swAutomatically configure Claude Code best practices for new projects: directory skeleton, AGENTS.md, hooks, and interactive placeholders fill. Invoke via `/init-project`.
Guides project setup through 6 phases: detects tech stack from package.json/requirements.txt/etc., creates CLAUDE.md, configures MCP memory and auto-loop hooks. For new/existing Claude Code projects.