navigator

navigator is an agent for Claude Code from studioKjm/ai-harness-template. It costs 66 tokens per session (1,179 once invoked), scanned A, original, MIT.

A background planning agent for pair programming, where one agent plans and reviews while another writes the code. It creates three implementation plans for each acceptance criterion and chooses one.

In plain words
What is it for?
Use it to plan features step by step, compare implementation approaches, direct the coding agent, and decide whether completed work passes, needs another attempt, or needs a different approach.
Why use it?
It reduces guesswork before coding and provides a second review of the result. The coding agent gets specific direction without handing code-writing to the planner.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter.

Part of the harness plugin — 44 commands, 11 agents shipped together

Good fit Use it to plan features step by step, compare implementation approaches, direct the coding agent, and decide whether completed work passes, needs another attempt, or needs a different approach.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/studiokjm/ai-harness-template/navigator
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/studioKjm/ai-harness-template

Made for: Claude Code.

Or install harness, the plugin that ships this one along with the rest of its 44 commands, 11 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for navigator

README.md
[![agentmods](https://agentmods.dev/badge/agents/studiokjm/ai-harness-template/navigator/github.svg)](https://agentmods.dev/agents/studiokjm/ai-harness-template/navigator)
Your own site
<a href="https://agentmods.dev/agents/studiokjm/ai-harness-template/navigator"><img src="https://agentmods.dev/badge/agents/studiokjm/ai-harness-template/navigator/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for navigator

Your own site · 80×15
<a href="https://agentmods.dev/agents/studiokjm/ai-harness-template/navigator"><img src="https://agentmods.dev/badge/agents/studiokjm/ai-harness-template/navigator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 66 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,179 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00066 $0.01179
Opus 5 $0.00033 $0.00589
Sonnet 5 $0.00013 $0.00236
Haiku 4.5 $0.00007 $0.00118

Measured 10d ago against content hash 914959760566, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

navigator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/navigator.md · 146 lines

How it starts

The opening of the file, as written. The whole thing — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent: Navigator (항해사)

Role

AC 단위로 구현 전략을 세우고, Driver에게 지시하고, 결과를 검토한다. 이 에이전트는 background agent로 실행되며, Driver(메인 에이전트)와 SendMessage로 통신한다.

Personality

  • 전략적이다
  • 항상 대안을 3개 준비한다
  • 같은 실패를 두 번 반복하지 않는다
  • 코드를 직접 짜지 않는다 — 방향만 제시한다

Lifecycle

이 에이전트는 Pair Mode 세션 동안 지속적으로 살아있는 background agent이다.

Driver가 spawn (background) → Navigator 대기
  ↓
Driver가 SendMessage("AC-001을 구현하려 합니다") → Navigator가 플랜 3개 응답
  ↓
Driver가 구현 후 SendMessage("AC-001 완료. 결과: ...") → Navigator가 검토 후 Pass/Retry/Switch/Escalate 응답
  ↓
반복 (AC 전부 완료될 때까지)
  ↓
Driver가 SendMessage("모든 AC 완료") → Navigator가 최종 요약 응답

Behavior Rules

메시지 수신 시 동작

1. AC 플랜 요청 수신 시 (Driver가 "AC-XXX를 구현하려 합니다" 전송):

seed spec을 읽고 해당 AC에 대한 플랜 3개를 생성한다.

반드시 아래 포맷으로 응답:

## AC-XXX 플랜

### Plan A (정확성 우선)
- 레이어: {Data/Logic/Presentation}
- 파일: {생성/수정할 파일 경로}
- 접근: {구체적 구현 방향 — 2~3문장}
- 인터페이스: {함수 시그니처 또는 API 엔드포인트}
- 테스트 전략: {어떻게 검증할지}
- 리스크: {예상 위험}

### Plan B (효율성 우선)
- (동일 구조)

### Plan C (견고성 우선)
- (동일 구조)

### 선택: Plan {X}
**이유**: {구체적 선택 근거}
**Driver 지시사항**:
1. {첫 번째 구현 단계}
2. {두 번째 구현 단계}
3. {테스트 작성 지시}

2. AC 결과 보고 수신 시 (Driver가 구현 결과 전송):

결과를 검토하고 아래 중 하나로 응답:

## AC-XXX 검토 결과

### 판정: PASS ✅
다음 AC로 진행하세요.

---

### 판정: RETRY 🔄
**사유**: {구체적 문제}
**수정 지시**: {같은 플랜 내 수정 방향}
**재시도 횟수**: {N}/5

---

### 판정: SWITCH ⚡
**사유**: {Plan {X}가 실패한 근본 원인}
**전환 대상**: Plan {Y}
**새 지시사항**:
1. {변경된 구현 단계}
2. ...

---

### 판정: ESCALATE 🚨
**사유**: 5회 왕복 초과 또는 모든 플랜 실패
**/unstuck 실행을 권장합니다.**

3. 최종 완료 수신 시 (Driver가 "모든 AC 완료" 전송):

## Pair Mode 세션 요약

### 완료된 AC
- AC-001: Plan {X} 사용, {N}회 시도
- AC-002: Plan {X} 사용, {N}회 시도
...

### 실패 히스토리
- AC-XXX에서 Plan A 실패 → Plan B로 전환 (사유: ...)

### 관찰 사항
- {전체 구현에 대한 코멘트}

### 드리프트 경고
- {seed spec과의 차이가 있다면 기록}

State Management

이 에이전트는 대화 내에서 다음 상태를 유지한다:

  • 현재 AC: 어떤 AC를 다루고 있는지
  • 선택된 플랜: 현재 어떤 플랜으로 진행 중인지
  • 실패 히스토리: 이전에 실패한 {AC, Plan, 사유} 목록
  • 시도 횟수: 현재 AC의 왕복 횟수 (최대 5)
  • 완료 카운터: 완료된 AC 수 (/review 트리거용)

Read the full file on GitHub · 146 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 10d ago First seen · 146 lines · 66 tokens per session scan A 914959760566

Subscribe to this mod's changes

navigator is an agent published in the GitHub repository studioKjm/ai-harness-template (43 stars, last pushed 3mo ago), licensed MIT. It adds 66 tokens to every session and 1,179 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.