---
name: Agent Consultant
slug: agent-consultant
category: AI Engineering
description: Agent Consultant gives technical advice or review on an approach, idea, or change. It is used for lightweight consultation without creating tasks or tracking harness work.
github: "https://github.com/enmanuelmag/agent-harness-kit/tree/main/.agents/skills/ahk-consultant"
language: TypeScript
stars: 180
forks: 9
install: "npx degit https://github.com/enmanuelmag/agent-harness-kit/tree/main/.agents/skills/ahk-consultant ~/.claude/skills/ahk-consultant"
installs_to: ~/.claude/skills/ahk-consultant
source_path: .agents/skills/ahk-consultant/SKILL.md
collection_size: 14
category_size: 3475
collection_url: "https://dirskills.com/collections/enmanuelmag/agent-harness-kit"
added: 2026-09-07T05:20:22.822Z
last_synced: 2026-09-07T05:20:22.822Z
canonical_url: "https://dirskills.com/skills/agent-consultant"
---

# Agent Consultant

Agent Consultant gives technical advice or review on an approach, idea, or change. It is used for lightweight consultation without creating tasks or tracking harness work.

**Install:**

```bash
npx degit https://github.com/enmanuelmag/agent-harness-kit/tree/main/.agents/skills/ahk-consultant ~/.claude/skills/ahk-consultant
```

## README

## Provider Delegation Guidance

- Sequential: Delegate only bounded, independent work; retain decisions and the final synthesis in the parent thread.
- Parallel: Delegate independent work in parallel when useful, then wait for every delegated result before continuing.
- Context Transfer: Give every delegated task a self-contained objective, scope, relevant context, restrictions, and output contract.
- Wait For Completion: Wait for the delegated result, then consolidate its findings in the parent thread.
- Inspect Progress: In the interactive CLI, use /agent to inspect delegated threads when needed.


You are in **lightweight consultation mode**. The user's topic: $ARGUMENTS

## Rules
- NO MCP calls — no tasks.*, no actions.*
- NO health.sh
- NO builder, NO reviewer
- Create files ONLY if user explicitly asked to save the output

## Process

1. Invoke **Explorer** as a subagent:
   > "Read-only context gathering — no MCP harness. The user wants technical advice on: `$ARGUMENTS`. Map the relevant codebase — existing patterns, current implementation, relevant dependencies. Return findings as plain structured text."

2. Invoke **Consultant** as a subagent in direct consultation mode, passing Explorer's output:
   > "Direct consultation — no MCP harness, no task. User's topic: `$ARGUMENTS`. Explorer found: [explorer output]. Provide structured technical advice. The provider has already injected available skills into your context — check what skills you are aware of, identify which are relevant to this topic, and reference them explicitly in your advice. If no installed skills seem relevant to the topic, recommend running `npx autoskills` to fetch appropriate skill packs."

3. Synthesize a final response for the user.

## Output structure
- **Context** — brief summary of relevant codebase state
- **Advice** — consultant's structured recommendations
- **Relevant skills** — skills from your context that apply, OR `npx autoskills` recommendation if none match
- **Risks & warnings** — what could go wrong with the proposed approach