A Claude Code / Cursor SDLC pipeline that takes a product brief and fully implements it — requirements, architecture, task planning, implementation, verification, and a handoff summary — with no manual intervention.

• Node.js 18+
• Claude Code or Cursor installed and authenticated
• A git repository to run the pipeline in

Each install also drops the walkthrough desktop app (orchestrate-walkthrough) onto your $PATH so the orchestrate skill can auto-launch it. Skip it with SKIP_APP=1.

curl -fsSL https://raw.githubusercontent.com/zolem/cc-sdlc/main/install.sh | bash

Installs the cc-sdlc Cursor plugin into ~/.cursor/plugins/local/cc-sdlc/ and the walkthrough app. Restart Cursor afterward.

curl -fsSL https://raw.githubusercontent.com/zolem/cc-sdlc/main/install.sh | MODE=claude bash

Installs the walkthrough app, then prints next steps. Inside Claude Code:

/plugin marketplace add zolem/cc-sdlc
/plugin install cc-sdlc@cc-sdlc

# Pin a version
curl -fsSL https://raw.githubusercontent.com/zolem/cc-sdlc/main/install.sh | VERSION=v1.1.0 bash

# Skip the desktop app
curl -fsSL https://raw.githubusercontent.com/zolem/cc-sdlc/main/install.sh | SKIP_APP=1 bash

# Install from your fork
curl -fsSL .../install.sh | REPO=my-org/cc-sdlc bash

Open Claude Code or Cursor in any repository and run:

/orchestrate <your product brief>

Example:

/orchestrate Build a URL shortener where users paste a long URL and get a short
link. Clicking the short link redirects to the original. Track click counts.
No user accounts required.

The pipeline runs fully automated from there.

If you're not sure what to put in the brief, use /generate-brief first. It walks you through a short Q&A and produces a structured brief you can then pass to /orchestrate.

/generate-brief task management app

Phase 1  Requirements      Product brief → user stories, personas, acceptance criteria
Phase 2  Context           Requirements + repo docs/code → focused context briefing
Phase 3  Architecture      Requirements + context → implementation plan + test cases (parallel)
Phase 4  Planning          All docs → ordered phase files with atomic work items
Phase 5  Implementation    Per phase: engineer (atomic commits) → qa + code review + security in parallel → up to 3 in-place fix iterations → walkthrough
Phase 6  Integration       Manual browser test of the fully integrated feature
Phase 7  Handoff           Summary with per-phase walkthroughs as the review entry point

All pipeline artifacts are written to .orchestrate/{feature-slug}/ at the repo root by default. Override with the ORCHESTRATE_OUT_DIR environment variable.

.orchestrate/{feature-slug}/
  requirements.md       — user stories and acceptance criteria
  context.md            — curated context briefing for downstream agents
  architecture.md       — implementation plan and component design
  test-plan.md          — user-perspective test cases
  task-index.md         — ordered phase list with dependency summary
  stack.json            — per-phase status, commits, and verification metadata
  phases/
    phase-1.md          — cohesive phase specs for engineer agents
    phase-2.md
    ...
  verification/
    phase-N/
      qa-report.md           — per-phase test suite results and coverage
      code-review-report.md  — per-phase senior code review
      security-report.md     — per-phase security findings
    manual-test-report.md    — end-of-pipeline browser walkthrough (if Chrome extension enabled)
  walkthroughs/
    phase-1.md          — narrated per-commit review guide for the phase
    phase-2.md
    ...

app/ contains an optional Tauri desktop app for browsing per-phase walkthroughs as the pipeline runs. Build it with cd app && npm install && npm run tauri build. If orchestrate-walkthrough ends up on $PATH, the orchestrate skill will auto-launch it at the start of each run. See app/README.md.

The manual-tester agent walks through user stories in a real browser using Claude's built-in Chrome integration. To enable it:

1. Install the Claude in Chrome extension in Chrome or Edge (v1.0.36+)
2. Launch Claude Code with Chrome integration enabled: claude --chrome

The pipeline works without this — the manual-tester is skipped if the extension is not available. All other verification (QA, security) runs regardless.

Source files live in src/ — edit agents in src/agents/ and skills in src/skills/. Each source file uses combined frontmatter with claude: and cursor: subsections for tool-specific fields. After editing, run the build to regenerate the .claude/ and .cursor/ output directories:

npm run build