constraint-discovery

# Constraint Discovery Skill

Spawn 5 parallel Explore agents to discover constraints that limit Claude's degrees of freedom.

## When to Use

- Invoked by `/ralph:start` Step 1.4.5 via Skill tool
- User asks to analyze project constraints
- User mentions "degrees of freedom" or "constraint scan"
- Standalone constraint analysis needed

## Agents

### Agent 1: Project Memory & Philosophy Constraints

```
Task tool parameters:
  description: "Analyze project memory constraints"
  subagent_type: "Explore"
  run_in_background: true
  prompt: |
    DEEP DIVE into project memory files AND FOLLOW ALL @ LINKS to discover constraints.

    STEP 1 - READ THESE FILES FIRST:
    - CLAUDE.md (project instructions, philosophy, forbidden patterns)
    - .claude/ directory (memories, settings, agents/*.md)
    - .claude/agents/*.md (agent definitions with @ references)
    - ROADMAP.md (P0/P1 priorities, explicit scope limits)
    - docs/adr/ (Architecture Decision Records)

    STEP 2 - FOLLOW ALL @ LINKS (UNLIMITED DEPTH):
    Parse each file for @ link patterns:
    - @path/to/file.md (relative to project root)
    - @ai_context/PHILOSOPHY.md (ai_context directory)
    - @projectname/path/to/file.md (project prefix)
    - @AGENTS.md, @README.md (root files)

    For EACH @ link found:
    1. Read the linked file
    2. Parse it for more @ links
    3. Recursively follow until no new @ links found

    STEP 3 - EXTRACT CONSTRAINTS FROM ALL FILES:
    - "Do NOT modify X" instructions
    - Philosophy rules (e.g., "prefer simplicity over features")
    - Explicit forbidden patterns
    - Scope limits from ROADMAP

    Return NDJSON: {"source":"agent-memory","severity":"CRITICAL|HIGH|MEDIUM","description":"...","file":"...","linked_from":"...","recommendation":"Ralph should avoid..."}
```

### Agent 2: Architecture & Coupling Constraints

```
Task tool parameters:
  description: "Analyze architectural constraints"
  subagent_type: "Explore"
  run_in_background: true
  prompt: |