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.
npx skills add yohey-w/kagemusha --skill advisor-gategit clone --depth 1 https://github.com/yohey-w/kagemushaWrote 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.
[](https://agentmods.dev/skills/yohey-w/kagemusha/advisor-gate)<a href="https://agentmods.dev/skills/yohey-w/kagemusha/advisor-gate"><img src="https://agentmods.dev/badge/skills/yohey-w/kagemusha/advisor-gate.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00145 | $0.15262 |
| Opus 5 | $0.00072 | $0.07631 |
| Sonnet 5 | $0.00029 | $0.03052 |
| Haiku 4.5 | $0.00015 | $0.01526 |
Grade A, and why
advisor-gate 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 7d 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.
How it starts
The opening of the file, as written. The whole thing — 1,035 lines — stays where its author put it; the contents beside it link to each section on GitHub.
advisor-gate
外部の高品質モデルへの諮問プロンプト生成・検収手順
English abstract — This is a procedure, not a prompt library. Before writing a consultation prompt for a frontier model, the agent must (1) fix a decision contract (what will change as a result of the answer), (2) enumerate decision pivots — facts whose alternative values would flip the conclusion — and classify each as PRESENT / BOUNDED / UNKNOWN / REDACTED / NOT-MATERIAL, (3) judge sufficiency as A (decidable) / B (conditional) / C (do not generate a prompt), and only then fill the standard template. It deliberately replaces "give MECE context", which has no completion test, with a procedural criterion that does. After the answer arrives it is stored verbatim, converted into a findings ledger keyed by (target × failure mechanism × impact) so later consultations cannot re-earn the same point, and gated by decision severity. Written mainly in Japanese; the structure is language-independent.
置き場所のプレースホルダ — 本文中の保存先は次の記号で書く。自分の環境の実パスへ読み替えること。
| 記号 | 意味 | 例 |
|---|---|---|
<state_dir>/ |
実行時の状態・作業ファイル置き場 | local/state/ |
<state_dir>/advisor_answers/ |
外部AI回答の逐語と検収カード | local/state/advisor_answers/ |
<state_dir>/advisor_answers/findings_ledger.yaml |
指摘台帳(§5.2のスキーマ) | 同上 |
<decision_journal> |
採用裁定の追記先(追記のみ) | judgment/decisions_journal.md |
置き場所の規約は3つだけ。①回答は逐語で1ファイル(加筆・整形・要約をしない)②検収カードは回答ファイルの末尾に併記(逐語と検収を1枚で開く)③指摘台帳は回答とは別ファイル(次回はこちらを渡し、回答全文は渡さない)。
目的
外部の高品質モデル(frontier級・拡張思考を持つ最上位の推論モデル)へ手で貼り付ける/APIで投げる「外部諮問プロンプト」を、AIエージェントが一貫した手順で生成し、返ってきた回答を検収する。
対象は、システム設計レビュー、文書批評、値付け、商談判断、発信戦略、キャリア相談などを問わない。 宛先を選ばない——拡張思考を持つ最上位モデルであれば、どのベンダーのものでも成り立つように書いてある(宛先固有の設定は、自分の環境側に1箇所だけ書く。§7を見よ)。
使う線引き: これは「日常の作業モデルへの発注」ではなく、一発で最大限の頭を使いたい重要な判断のレビューのための型である。安い作業モデルへ細かい仕事を投げるときにこの手順を回すのは過剰——そちらは普通の作業指示でよい。
本手順書が最適化するものは、次の4点である。
- 回答が具体的で、そのまま実装・修正・意思決定へ接続できること。
- オーナーの仮説や希望への自動追従を避けること。
- 渡された対象の内側だけでなく、フレーム外・境界面・未提示領域の穴を指摘できること。
- 継続諮問で同じ指摘を言い換えて繰り返させないこと。
外部AIの回答は原文のまま保存する。ただし、外部AIの発言だけを根拠に「事実」へ昇格させてはならない。回答は、事実ではなく「外部レビュー結果」「提案」「仮説」「リスク指摘」のいずれかとして扱い、採用裁定と実測結果を別に記録する。
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.
- 7d ago First seen · 1,035 lines · 145 tokens per session scan A 5afe07452527
advisor-gate is a skill published in the GitHub repository yohey-w/kagemusha (6 stars, last pushed yesterday), licensed MIT. It adds 145 tokens to every session and 15,262 once invoked, about $0.0007 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 2026-08-31.
Other skills, from other repositories
loop-builder
A design workbench for turning a vague task into a controlled, repeatable Agent Loop. An Agent Loop is a task process with defined steps, checks, stopping points, and recovery actions.
ui-visual-match-replicator
A screenshot-driven workflow for rebuilding a webpage or frontend so its rendered appearance matches a supplied visual. It compares screenshots, identifies the largest differences, and repeats the implementation and correction cycle.
akg-agents
An agent workflow for developing, generating, testing, and tuning AKG operators. AKG is a toolkit for generating optimized computational kernels.
openJiuwen-DeepSearch
A deep-research tool that plans searches, gathers and checks information, and produces research reports in Markdown, DOC, or HTML format. It coordinates several AI agents for complex investigations.
my-custom-skill
Example custom agent skill — a template to get started.
economic-reasoning
Economic decision-making for agents in EconomySpace. Activate when the agent observes currency, prices, income, tax, or trading opportunities to make informed financial decisions.