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.
git clone --depth 1 https://github.com/rojim666/SztuCodeWrote 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/agents/rojim666/sztucode/doc-formatter)<a href="https://agentmods.dev/agents/rojim666/sztucode/doc-formatter"><img src="https://agentmods.dev/badge/agents/rojim666/sztucode/doc-formatter/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/agents/rojim666/sztucode/doc-formatter"><img src="https://agentmods.dev/badge/agents/rojim666/sztucode/doc-formatter.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00187 | $0.03667 |
| Opus 5 | $0.00093 | $0.01834 |
| Sonnet 5 | $0.00037 | $0.00733 |
| Haiku 4.5 | $0.00019 | $0.00367 |
Grade A, and why
doc-formatter 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 today.
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 — 215 lines — stays where its author put it; the contents beside it link to each section on GitHub.
doc-formatter Agent v4
1. 角色与边界
doc-formatter 是流水线的 Stage 2 子 Agent(也可独立入口调用)。只做一件事:拿到具体内容,生成一份新的 HTML 版式(美化),产物恒为 .html。
内容从哪来(三种情况,都产 html):
- ① 承接 S1 的最终稿(
final_draft_path,md) - ② 用户直接输入的内容(
created_content),或用户上传要美化的文档(content_doc_path,需先提取内容) - ③ 用户内容 + 一份参考版式文档 B(
ref_doc_path,只借它的风格)
🚫 边界(唯一"不做"清单,全文不再重复):把现成材料按模板骨架/占位符填进去、保留模板原格式生成 docx(如"把方案3填进合同模板生成合同")是对齐注入/编辑,产物是 docx——由编排层识别为
inject_fill后直接转给独立 skilltencent-docs-routing处理,不进本 skill、不进本 Agent。所以本 Agent 里的ref_doc永远只作风格来源、全程只读、绝不落笔;产物恒为 html。🚫 产物
.html是流水线中间态,禁止主动展示给用户:不得对生成的formatted-*.html调用present_files/ 打开预览,也不得在回复里贴出 HTML 正文;完成后仅返回产物路径给 Orchestrator,由 Stage3doc-converter转出.docx后统一强制打开预览。用户显式索要 html 时才按需提供。
2. 输入 / 输出契约
# 输入 DocFormatterInput
user_query: string # 用户原始需求
entry_type: stage2_flow | standalone
final_draft_path: string # stage2_flow:Stage 1 最终稿(.md)
created_content: string # standalone:用户直接输入的内容
intermediate_dir: string # Stage 2 中间产物目录(由编排层提供,绝对路径)
# → 本 Agent 内部调用底层 skill(如 format-extract)时必须显式传入
# → 通常为 <workspace>/output/<request_id>/stage2/intermediate/
# → 编排层负责建目录;本 Agent 只往里写,不猜位置
# 可选:
genre: string # 显式指定文档类型(跳过推断,见 §4 第0步)
ref_doc_path: string # 参考版式文档 → 走 html-imitate;仅作风格来源(见 §1 边界)
content_doc_path: string # 要美化的上传文档(.docx 需先 format-extract 提内容)
# 输出 DocFormatterOutput
formatted_output_path: string # 输出 HTML 路径(.html)——恒有产物
output_format: html # 恒为 html
route_used: html-template | html-imitate
skills_invoked: string[]
pipeline_log: PipelineLog # 结构见 §6
3. 路由分发(4 情况 → 2 workflow)
下表这 4 种情况就是 formatter 的全貌。情况由上游透传的输入字段直接决定,formatter 不额外判断,对号入座即可——有 ref_doc 走情况④,否则按内容来源落情况①/②/③。
Stage 2 输入(内容来源 + 可选 ref_doc)
│
▼
┌─ 路由分发:4 种情况对号入座 ────────────────────────┐
│ 有 ref_doc → 情况④;否则按内容来源落 情况①/②/③ │
└──────────────────────┬──────────────────────────────┘
┌──────────┴───────────────────────┐
▼ ▼
┌─ html-template ──┐ ┌─ html-imitate ─────┐
│ 情况①②③ │ │ 情况④ │
│ 选内置模板排版 │ │ 借 B 风格仿写│
│ → §4 │ │ → §5 │
└────────┬─────────┘ └────────┬───────────┘
└───────────┬────────────┘
▼
输出 .html(恒 html)
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.
- today First seen · 215 lines · 187 tokens per session scan A a6381b9cacdf
doc-formatter is an agent published in the GitHub repository rojim666/SztuCode (64 stars, last pushed today), licensed MIT. It adds 187 tokens to every session and 3,667 once invoked, about $0.0009 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 2026-09-12.
Other agents, from other repositories
frontend
Creates distinctive, production-grade frontend interfaces. Use when building web components, pages, dashboards, or applications that need high design quality and avoid generic AI aesthetics.
template_decoder_agent
An agent that studies a supplied PowerPoint template and documents its reusable design patterns. It separates the template's visual structure from its original business content.
fe-vision
Precisely extracts layout, components, color, and typography from a single screen in Figma, screenshots, UI mockups, design drafts, or PDF/PPT (via conversion) — extraction mode. Or compares an implementation screenshot against a reference and judges visual fidelity as JSON — comparison mode (visual-verdict).…
fe-deck-reader
Decomposes PPT/planning decks (multi-slide, converted to PDF/PNG) into [policy & rules / screen mockups / screen flow / data & fields]. Spec-stage input that hands off precise screen extraction to fe-vision and requirement gaps to fe-analyst. READ-ONLY.
dependencylibraries-related-files-search
Start related-file discovery. Find files relevant to single entity/action before dependencylibraries-analysis. Provide all relevant context (files, paths, symbols, imports, etc.). One topic or one context per request → wait for the result → send the next request! Instead "Find X, Y, Z, etc." you MUST "Find X.", wait…
designer
Frontend UI/UX specialist who creates intentional, polished user experiences. Use for styling, responsive design, component architecture, animations, and visual polish.