scs-mcp 0.5.3

SCS - AI Capability Compiler. Turn scattered skills into verified, executable pipelines.

SCS — Skill Compiler System

中文版

Turn scattered skills into reliable workflows

You've collected many skill files. But when it's time to actually use them:

• "Which ones do I need for this task?"
• "Can they chain together? Or will it break halfway?"
• "Why do I have to manually piece them together, with no way to know if it will work?"

The core problem: Skills are defined separately, but they need to work together. Nobody builds the bridges.

---

What SCS Does for You

SCS is a compiler. It doesn't execute your skills — instead, it:

1. Scans all skills across all agents — automatically discovers skills from Claude Code, Cursor, Windsurf, Cline, and 40+ platforms
2. Connects them — automatically links skills that can work together
3. Verifies feasibility — tells you if the workflow will run, where gaps exist
4. Generates a plan — gives you a verified execution blueprint you can trust

Think of it like Terraform Plan for AI skills: you see a full preview with verification before anything runs.

---

Cross-Platform Skill Discovery

When you call scs_init() without specifying a path, SCS automatically scans all installed agent platforms:

scs_init()
→ Scanning 4 platforms: claude_code, cursor, windsurf, cline
→ Found 2465 skills, 176 artifact types

Important: SCS discovers skills from all platforms, but each agent can only invoke skills in its own skills directory. If the recommended skill comes from another platform, you'll need to:

• Install that agent, or
• Copy the skill to your current agent's skills directory

The execution plan shows which platform each skill comes from, so you know what's available.

---

One-Click Install

After cloning or downloading SCS, run one command:

# Install to Claude Code
python -m scs.cli install --agent claude_code --force

# Or install to all supported agents
python -m scs.cli install --all --force

Restart your agent and SCS tools are ready.

What gets installed:

1. MCP tools — 8 SCS tools available in your agent
2. SCS Router Skill — automatically injected to your agent's skills directory, guiding when to use SCS

Supported agents: Claude Code, Cursor, Windsurf, Cline

---

Manual Install

If you prefer manual configuration, add to ~/.claude/settings.json:

{
  "mcpServers": {
    "scs": {
      "command": "python",
      "args": ["-m", "scs.mcp_server"],
      "cwd": "/path/to/SCS",
      "env": { "PYTHONUTF8": "1" }
    }
  }
}

Replace /path/to/SCS with your SCS directory.

---

CLI Commands

Command	What it does
scs install --agent claude_code	Install to Claude Code
scs install --all --force	Install to all supported agents
scs config	Show configuration help
scs path	Show installation path

---

Tool Overview

After installation, your agent will have 8 new tools. You mainly use these 5:

scs_init — Initialize (call this first)

Scans skills, connects them, generates execution plans.

# Scan all agents' skills
scs_init()

# Solve optimal path to a target
scs_init(target="ValidatedPRD", prefer="balanced")

scs_scan — Scan (for large skill sets)

Scans and registers skills, but doesn't generate plans. Use when you have many skills.

scs_scan()
→ Scanned and registered 2465 skills

scs_compile — Compile

Generates execution plans from already-scanned skills.

scs_compile(target="ValidatedPRD")
→ Compiled, found optimal path

scs_list — List things

scs_list(type="dag")       # All execution plans
scs_list(type="skills")    # All skills
scs_list(type="artifacts") # All artifact types

scs_get — Get details

scs_get(type="dag", id="plan_xxx")
scs_get(type="skill", id="prd_generator")

---

Typical Workflow

# 1. Start fresh
scs_clear()

# 2. Scan and generate plans
scs_init()

# 3. See what was created
scs_list(type="dag")

# 4. Find optimal path to your target
scs_init(target="ValidatedPRD", prefer="balanced")

# 5. Understand why SCS chose this path
scs_diagnostics(plan_id="plan_xxx")

---

Large Skill Set Workflow

When you have 2000+ skills, use the phased approach to avoid timeout:

scs_scan()                       # Step 1: Scan (~90s)
scs_list(type="artifacts")       # Step 2: Explore
scs_compile(target="YourTarget") # Step 3: Compile (~30s)

---

Use Cases

SCS covers 8 industries, 176 artifact types:

Industry	Example artifacts
Software	Requirements doc, source code, test report
Legal	Contract, legal opinion, due diligence report
Marketing	Marketing plan, brand guidelines, ad copy
Finance	Financial report, invoice, ROI analysis
HR	Job description, resume, performance review
Research	Research paper, literature review, grant application
Content	Article, blog post, video script
Operations	Service ticket, SOP, incident report

---

SCS — Compile skills, not chaos.