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 u-ichi/reviewable-html-workbench --skill reviewable-design-docgit clone --depth 1 https://github.com/u-ichi/reviewable-html-workbenchWrote 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/u-ichi/reviewable-html-workbench/reviewable-design-doc)<a href="https://agentmods.dev/skills/u-ichi/reviewable-html-workbench/reviewable-design-doc"><img src="https://agentmods.dev/badge/skills/u-ichi/reviewable-html-workbench/reviewable-design-doc/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.
<a href="https://agentmods.dev/skills/u-ichi/reviewable-html-workbench/reviewable-design-doc"><img src="https://agentmods.dev/badge/skills/u-ichi/reviewable-html-workbench/reviewable-design-doc.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 5 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Prompt Injection · line 52 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
- high Prompt Injection · line 135 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
- high Prompt Injection · line 244 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
- medium Agent Snooping · line 426 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
- medium Agent Snooping · line 520 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
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 | $0.00236 | $0.12811 |
| Opus 5 | $0.00118 | $0.06406 |
| Sonnet 5 | $0.00047 | $0.02562 |
| Haiku 4.5 | $0.00024 | $0.01281 |
Grade A, and why
reviewable-design-doc 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.
How it starts
The opening of the file, as written. The whole thing — 580 lines — stays where its author put it; the contents beside it link to each section on GitHub.
reviewable-design-doc
役割
設計資料としてレビューできる構造を作り、最終HTML生成は visual-html-renderer に渡す。
レビュー完了後は annotations/comments.json を読み、明確な指摘は設計へ反映し、確認・回答が必要な指摘は add-reply CLI でHTMLの同じコメントスレッドへ書き戻す（チャットへの回答ではなくHTML上の返信として）。
Role
Create a design document that can be reviewed in the browser. This skill owns design structure, review intent, comment ingestion, and comment-thread replies. Final HTML rendering is delegated to visual-html-renderer. After review, read annotations/comments.json, apply clear resolved feedback, and write clarification replies back into the same HTML comment thread with add-reply.
言語方針 / Language behavior
Follow the language of the latest user request for progress updates, final responses, and review handoff text. レビューコメントへの返信は、原則としてそのコメント本文の言語に合わせる。日本語コメントには日本語で、英語コメントには英語で返信する。設計本文や引用内容は、ユーザーが翻訳を求めない限り勝手に翻訳しない。
基本手順
- 設計対象、読者、レビュー目的、完了条件を整理する。
- 要求、制約、アーキテクチャ、代替案、意思決定、未決事項へ分解する。
- レビュー観点とコメントしてほしい範囲を明示する。
- 最初から
document-model.jsonを作る。設計本文の下書きや中間成果物として.mdを作らない。 - 文書モデルには、要求、制約、アーキテクチャ、代替案、意思決定、未決事項、レビュー観点を、それぞれ現行rendererで有効なHTML表現を選んだblockとして入れる。
image.generation_status=requestedのブロックがある場合は、imagegenskillで画像を生成し、attach-imageCLIで文書モデルへ添付する。check-modelCLIで最終render前の文書モデル品質を検査する。renderCLIでHTML bundleを生成する。validateCLIでHTML bundleを検証する。- ユーザー向け最終HTMLでは既定で
previewCLIを--mode autoで起動し、返却JSONのurlとstop_commandを最終応答に必ず書く。 - preview 起動直後に、Monitor ツールで
watch-commentsを開始する。これによりブラウザからのコメントを自動検知できるようになる。Monitor 起動コマンド:python3 -m scripts.html_review_workbench.cli watch-comments --root <output-dir>。イベント受信後の処理は「コメント自動回答と解決待ちゲート」セクションに従う。 - ユーザーがコメントを入れたら「レビューコメントへの対応」セクションに従う。
Basic Workflow
- Clarify the design target, audience, review purpose, and completion criteria.
- Split the material into requirements, constraints, architecture, alternatives, decisions, unresolved issues, and review points.
- Create
document-model.jsonfrom the start; do not create a.mddraft as the design body. - Choose renderer-supported HTML blocks for each design unit.
- Generate requested images with
imagegenand attach them before rendering. - Run
check-model,render,validate, andpreview. - Start
watch-commentsafter preview startup. - When the user adds comments, ingest them, reply in the HTML thread to every thread listed in
needs_agent_review_threads, and apply resolved feedback only after gates allow it.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 10d ago First seen · 580 lines · 236 tokens per session scan A 18ccd4cc3146
reviewable-design-doc is a skill published in the GitHub repository u-ichi/reviewable-html-workbench (297 stars, last pushed 1mo ago), licensed MIT. It adds 236 tokens to every session and 12,811 once invoked, about $0.0012 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.
Other skills, from other repositories
reconcile
Cross-reference a paper against the collection. Finds which cited papers are already collected, which are new leads, which collection papers cite this one, and reconciles all cross-references bidirectionally. Run on a single paper directory or use --all for the entire collection.
author-context
Create a propstore context for one paper with CEL assumptions, parameters, and perspective. Every claim extracted from the paper will reference this context. Required before extract-claims if the paper does not already have one.
paper-retriever
Retrieve a scientific paper PDF given an arxiv URL, DOI, or paper title. Downloads to papers/ directory. Handles open-access retrieval (arxiv direct, Chrome print-to-pdf for publisher-direct HTML, Unpaywall, title-based open-repository search); for paywalled papers it hands off to whichever paper-retriever- access…
register-concepts
Register concepts from a paper into its propstore source branch using pks source propose-concept.
adjudicate
Systematically adjudicate disagreements across a paper collection. Produces ruthless verdicts on who was wrong, what supersedes what, and what the best current understanding is. Organized by topic clusters with actionable replacement values for implementation.
author-lifting-rules
Author DeLP/McCarthy context lifting rules between existing context YAMLs so cross-context claim reasoning can proceed. Collection-level, runs after all papers are ingested. Targets the CONTEXTPHINODE records that accumulate when multiple per-paper contexts share concepts with no authored bridges.