/opt/homebrew/Library/Homebrew/cmd/shellenv.sh: line 18: /bin/ps: Operation not permitted name: os-dev-knowledge-skill description: > Skill providing structured knowledge about OS 4.1 / Claude Code configuration patterns, safety constraints, and common orchestration pitfalls. Intended for os-dev-grand-architect, os-dev-architect, and OS-Dev specialists. LOCAL to claude-vibe-config repo only.

OS-Dev Knowledge Skill – OS 4.1 Configuration & Safety

NOTE: This skill is LOCAL to claude-vibe-config repo only.

This skill captures key knowledge for working on OS 4.1 / Claude Code config: lanes, commands, agents, skills, MCPs, hooks, and memory integration.

Use this skill when:

• Planning changes to orchestration lanes or phase configs.
• Adding or modifying commands/agents/skills for OS 2.x.
• Configuring MCPs, hooks, or memory behavior that affect all lanes.

Configuration Surfaces

• Lanes and pipelines:
  • docs/pipelines/*.md
  • docs/reference/phase-configs/*.yaml
• Agents and commands:
  • agents/*.md, agents/dev/*.md
  • commands/*.md
• Skills and MCPs:
  • skills/*/SKILL.md
  • mcp/* (MCP configs, server docs)
• Memory and orchestration:
  • .claude/memory/ (Workshop + code-index.db)
  • .claude/orchestration/ (phase_state, evidence, playbooks)

Core Principles

1. Orchestrators never implement
   • /orca-* and *-grand-architect agents only coordinate.
2. Specs before complex changes
   • Requirements specs in .claude/requirements/<id> are required for complex, global changes.
3. Memory-first
   • Workshop and code-index.db should be queried before expensive context.
4. Safety over convenience
   • Never default to dangerous flags or uncontrolled hooks.
5. Edit-not-rewrite
   • Minimal diffs in config files; respect existing patterns where safe.

Common Pitfalls

• Introducing CLI flags (like --dangerously-skip-permissions) as defaults.
• Adding hooks that run on every session and execute arbitrary shell commands.
• Scattering temporary logs and config fragments outside .claude/.
• Using YAML arrays for tools: in agents, causing silent tool failures.

Best Practices

• Keep OS-Dev agents and commands lightweight:
  • Focused scopes, minimal tool lists, clearly documented behaviors.
• Treat standards as evolving:
  • When recurring problems are found, codify them in OS-Dev standards and enforce via gates.
• Integrate Response Awareness:
  • Use RA tags to document assumptions and path choices, especially around safety and global behavior.

Agents and orchestrators should reference this skill when reasoning about OS-Dev changes to stay aligned with the overall OS 4.1 design philosophy.