Agent OrchestrationSAFE

Context management for Claude Code. Hooks maintain state via ledgers and handoffs. MCP execution without context pollution. Agent orchestration with isolated context windows.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: agent-orchestration
description: Agent Orchestration Rules
user-invocable: false
---

# Agent Orchestration Rules

When the user asks to implement something, use implementation agents to preserve main context.

## The Pattern

**Wrong - burns context:**
```
Main: Read files → Understand → Make edits → Report
      (2000+ tokens consumed in main context)
```

**Right - preserves context:**
```
Main: Spawn agent("implement X per plan")
      ↓
Agent: Reads files → Understands → Edits → Tests
      ↓
Main: Gets summary (~200 tokens)
```

## When to Use Agents

| Task Type | Use Agent? | Reason |
|-----------|------------|--------|
| Multi-file implementation | Yes | Agent handles complexity internally |
| Following a plan phase | Yes | Agent reads plan, implements |
| New feature with tests | Yes | Agent can run tests |
| Single-line fix | No | Faster to do directly |
| Quick config change | No | Overhead not worth it |

## Key Insight

Agents read their own context. Don't read files in main chat just to understand what to pass to an agent - give them the task and they figure it out.

## Example Prompt

```
Implement Phase 4: Outcome Marking Hook from the Artifact Index plan.

**Plan location:** thoughts/shared/plans/2025-12-24-artifact-index.md (search for "Phase 4")

**What to create:**
1. TypeScript hook
2. Shell wrapper
3. Python script
4. Register in settings.json

When done, provide a summary of files created and any issues.
```

## Trigger Words

When user says these, consider using an agent:
- "implement", "build", "create feature"
- "follow the plan", "do phase X"
- "use implementation agents"

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (1)

.claude/scripts/mcp

Gates applied: no_behavioural_pass.

[![ModelsAtlas audit: Agent Orchestration](https://modelsatlas.com/skills/parcadei/agent-orchestration/badge.svg)](https://modelsatlas.com/skills/parcadei/agent-orchestration/)

Not your listing, or something out of date? Claim or update your listing.