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 agentmods add skills/channinglua/prax-agent/knowledge-compilenpx skills add ChanningLua/prax-agent --skill knowledge-compilegit clone --depth 1 https://github.com/ChanningLua/prax-agentWhat 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 | $0.00040 | $0.01871 |
| Opus 5 | $0.00020 | $0.00936 |
| Sonnet 5 | $0.00008 | $0.00374 |
| Haiku 4.5 | $0.00004 | $0.00187 |
Grade A, and why
knowledge-compile 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 2d 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 — 189 lines — stays where its author put it; the contents beside it link to each section on GitHub.
知识编译技能:把 raw markdown 变 wiki
把一个目录下的散装 markdown(抓取的推文、知乎回答、文章笔记 ...)整理成结构化 wiki。产出在 Obsidian 里双链可用。
何时用
- 用户说"整理今天抓的内容"、"把 X 目录编译成 wiki"、"生成日报/周报"
- 和
browser-scrape技能配套:先抓后编 - 和
cron任务配套:定时 pipeline 的最后一环
输入
一个目录,里面可能有:
- 多个
*.md文件(来自不同来源:twitter、zhihu、bilibili、read...) - 可选的
raw/*.json(原始抓取数据,用作溯源)
产出(约定)
写回同一个目录下:
<input-dir>/
├── index.md # TOC + 一览表
├── daily-digest.md # 一屏内看完:一句话总览 + 100 字摘要 + 今日亮点
├── topics/
│ ├── <topic-slug>.md # 按主题聚合,内部用 Obsidian 双链 [[source-file]] 回引原文
│ └── ...
└── raw/ # 保留原始文件(若已有)
硬约束:
- 只新增 / 覆盖上面四类文件;不要删
.md原文或raw/ - 原文保留在原路径,通过
[[文件名]]双链引用 - 所有时间用 ISO 格式(
2026-04-22T17:05:00+08:00) - 中文优先,技术术语保留英文原词
工具选择(很关键):
- 创建新文件(
index.md、daily-digest.md、topics/<slug>.md):必须用Write工具,它会自动mkdir -p父目录。 - 修改已存在文件:用
Edit或HashlineEdit。 HashlineEdit/Edit对不存在的路径会报File not found—— 不要用它们写新文件。
标准执行流程
步骤 1:扫描输入目录
Glob: <input-dir>/*.md
Glob: <input-dir>/raw/*.json # 可选
读每个 .md,抽取 frontmatter(至少 source / url / scraped_at 可能存在),取第一段作为摘要候选。
步骤 2:聚类成主题
- 用内容相似性 + 关键词共现,把文件分成 3-7 个主题
- 主题名短、名词化、中文:如
模型发布、AI 工具、工程实践、安全与对齐 - 每篇文章归入一个主主题(避免重复计数带来的膨胀)
- 主题 slug 用拼音或英文短语:
model-releases、ai-tools、engineering、safety-alignment
步骤 3:写 topics/<slug>.md
模板:
---
topic: 模型发布
slug: model-releases
article_count: 5
generated_at: 2026-04-22T17:05:00+08:00
---
# 模型发布
## 本期概要
<2-3 句话说明本主题下的整体动态>
## 条目
### [[twitter-17xxx]] — Anthropic 发布 Claude Opus 4.7
- 来源:X @AnthropicAI
- 时间:2026-04-22 09:30
- 要点:
- <bullet 1>
- <bullet 2>
### [[zhihu-abc123]] — <标题>
...
关键点:
[[文件名]]不带.md后缀,指向原文件 stem- 一条一个 H3,不要堆成长段落
- 要点用 bullet,每条 ≤ 30 字,严禁复述原文
步骤 4:写 index.md
---
generated_at: 2026-04-22T17:05:00+08:00
article_count: 23
topic_count: 5
---
# 索引 · 2026-04-22
## 主题一览
| 主题 | 条目数 | 入口 |
|---|---|---|
| 模型发布 | 5 | [[topics/model-releases]] |
| AI 工具 | 8 | [[topics/ai-tools]] |
...
## 原始文件
- [[twitter-17xxx]]
- [[zhihu-abc123]]
- ...(按时间倒序)
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.
- 2d ago First seen · 189 lines · 40 tokens per session scan A 88501ecec9b8
knowledge-compile is a skill published in the GitHub repository ChanningLua/prax-agent (272 stars, last pushed 1mo ago), licensed MIT. It adds 40 tokens to every session and 1,871 once invoked, about $0.0002 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
knowledge-base-management
Obsidian 知识库全生命周期管理:三层架构、素材入库(ABC分级)、健康检查、GBrain/GraphRAG/LLM Wiki 三件套集成、目录整理.
knowledge_base
Manage the user's personal knowledge base — knowledge graph, documents, and wiki vault.
wiki-project
Create or update project pages in the Obsidian vault's 03 PROJECTS — status changes, decision log entries, milestones, meeting notes, future-project ideas. Use when the user says "log a decision for ", "update status", "new project page", "what's the state of ", "we had a meeting about ", "add a project idea", or…
obsidian-markdown
Obsidian Flavored Markdown syntax reference — wikilinks, embeds, callouts, properties/frontmatter, tags, comments, block IDs. Use when creating or editing .md files in an Obsidian vault, or when the user mentions wikilinks, callouts, frontmatter, embeds, or Obsidian notes. The wiki- skills assume this syntax.
obsidian
Operational skill for Claude to automate Obsidian vaults via URI schemes, Templater/Dataview patterns, Markdown structure, and local plugin workflows.
wiki-capture
Route anything into the Obsidian vault (second brain) — a thought, URL, concept, snippet, or the 00 INBOX backlog. Use when the user says "add to my wiki", "capture this", "note this down", "save to vault", "process my inbox" — and also on wiki-free phrasings like "jot this down", "remember this idea", "TIL"…