Harness Plan Brief logo

Harness Plan Brief

CommunityPopular
Chachamaru127
harness-plan-brief

Generate a Plan Brief HTML for non-engineer vibecoders before implementation starts. Searches harness-mem (project-only) for relevant past decisions, patterns, and Plans archive entries, then renders a single-file HTML artifact summarizing understanding, options, risks, acceptance criteria, and confidence. Use when the user requests a planning preview, a non-engineer-friendly summary before approval, or says: plan brief, planning preview, 計画概要, 計画レビュー. Do NOT load for: actual implementation, code review, release work.

Overview

PublisherChachamaru127
Repositoryclaude-code-harness
Skill nameharness-plan-brief
Stars
3.2K
Forks
302
Bundled files
1
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 1 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by Chachamaru127 on GitHub. Read the source before you install it.

Installation

Install the Harness Plan Brief AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/Chachamaru127/claude-code-harness/tree/main/codex/.codex/skills/harness-plan-brief
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git /tmp/claude-code-harness
mkdir -p .claude/skills
cp -r /tmp/claude-code-harness/codex/.codex/skills/harness-plan-brief .claude/skills/harness-plan-brief
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Harness Plan Brief in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Harness Plan Brief on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Harness Plan Brief is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

harness-plan-brief

非エンジニアの発注者・プロデューサー職向けに、Claude が着手しようとしている計画を HTML 1 枚 で提示するスキル。 発注者の認知負荷ピーク (1) 計画理解の段階で使う。

Quick Reference

  • 「Plan Brief を作って」 → このスキル
  • 「実装前にざっくり整理」 → このスキル
  • 「非エンジニア向けに計画を見せて」 → このスキル

責任境界

範囲このスキルの責務
検索現プロジェクトのみ (project: <current>, strict_project: true を必ず指定)
クロスプロジェクトやらない (Phase 65.3 以降で --cross-project-group <name> flag で opt-in 解放)
書き込みやらない (Plan Brief 承認後の memory write は plan-brief-record-decision.sh の責務)
plan_readiness 算出scripts/plan-brief-compile.sh に委譲。互換フィールド名 confidence は残すが、意味は DoD 明確度 + 依存解決率に限定

入力

引数 [task-description] にユーザーの request を渡す。 引数なしの場合は、会話の原依頼と選択済み計画から回収する。複数の候補が残って判断を変える場合だけ確認する。 計画の目的、担当範囲、未決事項と合格条件を示し、計画概要の作成依頼から実装を始めない。

出力

出力パス形式
Plan Brief HTML.claude/state/views/plan-brief-<timestamp>.html単独で開ける HTML (no server, no JS framework)
Plan Brief context JSON.claude/state/views/plan-brief-<timestamp>.context.jsonplan-brief-context.v1 schema

Schema: plan-brief-context.v1

json
{
  "schema": "plan-brief-context.v1",
  "user_request": "string (ユーザーの request 原文)",
  "my_understanding": "string (Claude の理解を 1-3 段落で)",
  "options": [
    { "name": "string", "summary": "string", "pros": ["string"], "cons": ["string"] }
  ],
  "risks": [
    { "kind": "string", "severity": "info|warn|critical", "description": "string", "mitigation": "string" }
  ],
  "acceptance_criteria": [
    { "id": "string", "description": "string", "verifiable_by": "string" }
  ],
  "tdd_required": "yes|no|skip:<reason>",
  "confidence": 0,
  "confidence_evidence": ["string (plan_readiness evidence: DoD clarity + dependency resolution only)"],
  "related_decisions": [
    { "id": "string", "title": "string", "relevance": "string" }
  ],
  "similar_past_plans": [
    { "archive_path": "string", "phase": "string", "outcome": "cc:完了|cc:WIP|cc:TODO|skipped", "relevance": "string" }
  ],
  "project": "string",
  "generated_at": "ISO8601"
}

完全 schema は schemas/plan-brief-context.v1.schema.json を参照。

Execution Flow

スキル起動時、Claude は以下の手順で動作する。

Step 1: project name を解決

bash
PROJECT_NAME="$(basename "$(git rev-parse --show-toplevel)")"

PROJECT_NAME が空 (git 外) の場合は current をデフォルトに使う。

Step 2: harness-mem を project-only で検索する (default)

引数に --cross-project-group <name> flag がない場合 (default behavior):

mcp__harness__harness_mem_search を 必ず 以下のパラメータで呼び出す:

project: <PROJECT_NAME>
strict_project: true
query: <user request>
expand_links: true
limit: 5

重要: project パラメータは必須。空文字列や null を渡してはならない。 strict_project: true を指定し、cross-project な検索は絶対に行わない。 必要なら tags filter で decision / pattern を絞ってもよいが、project は固定。

過去 decision (D1-D41) / pattern (P1-P33) / Plans archive 28 件から類似案件を最大 5 件取得する。

Step 2 (alt): cross-project search (Phase 65.3.5 opt-in)

引数に --cross-project-group <name> flag がある場合のみ:

D43 Option α (MCP N-call) に従い、以下の手順で cross-project 検索を行う。

bash
# (a) group → member projects に解決 (yaml SSOT)
MEMBERS_JSON="$(bash scripts/load-cross-project-groups.sh --group "<name>" 2>/dev/null)" || {
  echo "ERROR: cross-project group not found: <name>" >&2
  exit 1
}
# MEMBERS_JSON は ["proj1","proj2",...] 形式の JSON 配列

MEMBERS_JSON が [] (空配列) の場合は warning を出して default の単一 project search に fallback。

MEMBERS_JSON が非空の場合、各 member project に対して MCP search を 1 回ずつ発行 する:

for each project in MEMBERS_JSON:
  mcp__harness__harness_mem_search(
    project: <member>,
    strict_project: true,
    query: <user request>,
    expand_links: true,
    limit: 5
  )

各 search 結果を client 側でマージ・dedupe (id 単位)・relevance_score 降順 sort し、最大 5 件に絞る。 合計呼び出し数が多くなる (group が 5 project なら 5 回) ため、レイテンシは増える点に注意。

D43 判断 1 の根拠: MCP tool schema には projects: [array] も strict_project: false も exposed されていないため、横断検索は client 側 N-call が唯一の選択肢。 詳細は .claude/rules/cross-repo-handoff.md の「Phase 65.3 実装決定事項 (D43)」参照。

cross-project 結果には Layer 2/3 (Phase 65.3.2-65.3.4) の redaction を必ず通すこと:

  • HTML レンダリング時に bash scripts/render-html.sh ... --with-redaction を使用
  • これにより辞書 + NER + final scan の 3 段で固有名詞が漏れない

Step 3: context JSON を組み立てる

scripts/plan-brief-compile.sh を使って、mem search 結果から plan-brief-context.v1 schema 準拠の JSON を構築する。

Phase 105.3 以降、Plan Brief の confidence は後方互換のフィールド名であり、 表示上の意味は plan_readiness として扱う。算出軸は次の 2 つだけに固定する。

  • DoD 明確度: request / DoD に機械検証できる数値・条件がどれだけ含まれるか
  • 依存解決率: 類似 Plans のうち依存が完了済みとして扱えるものの割合

過去類似案件の成功率や関連 Decision / Pattern 件数は context-only の根拠として表示し、 readiness 点数へ別軸加算しない。これは「AI の理解度」「成功確率」と誤読されるのを避けるため。

options / risks / acceptance_criteria は常に 1 件以上生成する。 mem search が空でも、以下を最低限埋める。

  • options: 推奨案を 1 件以上。必要なら代替案を追加し、pros / cons を付ける
  • risks: readiness 誤読、scope creep、未観測データなど今回の計画固有リスクを 1 件以上
  • acceptance_criteria: 実行後に機械検証または目視確認できる条件を 1 件以上

例:

bash
jq -n \
  --arg req "$USER_REQUEST" \
  --arg proj "$PROJECT_NAME" \
  --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
  '{
    schema: "plan-brief-context.v1",
    user_request: $req,
    my_understanding: "(まだ未着手)",
    options: [{name:"Option A: 最小検証で進める", summary:"DoD と依存を先に確認してから実装", pros:["影響が小さい"], cons:["大きな再設計は別タスク化が必要"]}],
    risks: [{kind:"readiness-misread", severity:"warn", description:"plan_readiness を AI の理解度として誤読するリスク", mitigation:"DoD 明確度 + 依存解決率だけの指標として evidence に明記"}],
    acceptance_criteria: [{id:"AC-1", description:"Plan Brief context が非空の options / risks / acceptance_criteria を含む", verifiable_by:"tests/test-plan-brief-compile.sh"}],
    confidence: 0,
    confidence_evidence: ["plan_readiness DoD 明確度: 0/60", "plan_readiness 依存解決率: 0/40"],
    tdd_required: "no",
    related_decisions: [],
    similar_past_plans: [],
    project: $proj,
    generated_at: $ts
  }' > "$CONTEXT_JSON"

Step 4: HTML を生成する

scripts/render-html.sh (Phase 65.1.1) を templates/html/plan-brief.html.template で呼ぶ:

HTML には TDD 判定を 1 行で表示する。 形式は tdd_required: yes、tdd_required: no、または tdd_required: skip:<reason> のいずれかにする。

bash
bash scripts/render-html.sh \
  --template plan-brief \
  --data "$CONTEXT_JSON" \
  --out "$HTML_OUT"

diagram-design skill がインストールされていれば図の描画に使う。無ければ静的レイアウトのまま。

Step 5: ブラウザで自動 open する

scripts/plan-brief-open.sh で OS 別 dispatch:

bash
bash scripts/plan-brief-open.sh "$HTML_OUT"

BROWSER=true の env が設定されている場合 (CI 環境)、open は skip され printf で path だけ出力する。

Step 6: ユーザー承認待ち

「この理解で実装に進んでよいか」を確認する。 承認後の memory write は別スキル (Phase 65.1.4 の plan-brief-record-decision.sh) の責務。

失敗時の挙動

失敗挙動
mcp__harness__harness_mem_search 不達警告を表示し、related_decisions / similar_past_plans を空配列で続行
git rev-parse --show-toplevel 失敗PROJECT_NAME=current で続行
render-html.sh 失敗エラーを stderr に出力し exit 1
plan-brief-open.sh 失敗HTML path を stdout に出力するだけで exit 0 (browser open は best-effort)

Related

  • scripts/render-html.sh (Phase 65.1.1) — HTML テンプレートエンジン
  • scripts/plan-brief-compile.sh (Phase 65.1.3) — context compilation
  • scripts/plan-brief-record-decision.sh (Phase 65.1.4) — 承認 memory write
  • harness-accept skill (Phase 65.2.1) — 受け入れ判断スキル (対構造)
  • harness-progress skill (Phase 65.4.1) — 進行管理スキル (対構造)

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Harness Plan Brief AI skill do?

Generate a Plan Brief HTML for non-engineer vibecoders before implementation starts. Searches harness-mem (project-only) for relevant past decisions, patterns, and Plans archive entries, then renders a single-file HTML artifact summarizing understanding, options, risks, acceptance criteria, and confidence. Use when the user requests a planning preview, a non-engineer-friendly summary before approval, or says: plan brief, planning preview, 計画概要, 計画レビュー. Do NOT load for: actual implementation, code review, release work.

Why use Harness Plan Brief on TypingMind?

Because you install it once and use it with any model. Harness Plan Brief is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Harness Plan Brief in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Chachamaru127/claude-code-harness/tree/main/codex/.codex/skills/harness-plan-brief. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Harness Plan Brief?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Harness Plan Brief?

As many as you like. As long as a model supports skills, you can use Harness Plan Brief with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Harness Plan Brief AI skill free?

Yes. It is published on GitHub by Chachamaru127 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇