← Back to Skills Marketplace
gcdd1993

Digest Builder

by 阿晨聊技术 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
94
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install digest-builder
Description
将聚合池中的原始候选内容转换为可判断的 Digest：优先从 FreshRSS 的未读列表经 Google Reader API 拉取候选，再执行 URL 去重、相似事件聚类、正文抓取检查、噪音过滤、摘要生成与初步排序。用于需要把一批原始 feed 条目或 FreshRSS 未读条目整理成结构化候选文档的场景。
README (SKILL.md)

Digest Builder

以简体中文工作。目标是把原始候选整理成干净、结构化、适合快速判断的候选池，而不是直接产出日报终稿。若工作区已经配置 FreshRSS，则默认先从 FreshRSS API 拉取未读列表，再进入 Digest 预处理。

核心任务

  • 通过 FreshRSS Google Reader API 拉取未读列表
  • 读取配置中的 api_password
  • URL 精确去重
  • 相似事件聚类
  • 正文抓取可用性检查
  • 噪音过滤
  • 摘要生成
  • 标签补全
  • 初步排序

处理顺序

  1. 若已配置 FreshRSS，先通过 Google Reader API 完成登录并拉取未读列表。
  2. 默认按“昨天”这个自然日窗口切片未读条目；只有用户明确指定其他日期时才覆盖。
  3. 先做 URL 归一和精确去重。
  4. 再识别同一事件的相似条目，进行聚类或合并。
  5. 检查正文是否可抓取、是否存在明显质量问题。
  6. 剔除标题党、无正文、广告型、明显重复价值不足的内容。
  7. 为保留条目生成 2 到 5 行摘要。
  8. 补充主题标签、风险标签、行动价值标签。
  9. 按公共重要性、个人相关性、来源可靠度做初排序。

去重硬约束

  • Digest 必须严格去重，禁止同一事件、同一结论、同一新闻变体重复出现。
  • 去重分三层执行：
    • URL 层：相同链接、归一化后等价链接、追踪参数不同但正文相同的链接只保留一条。
    • 事件层：同一事件的多来源只保留信息增量最高的一条为主条目，其余仅作为参考来源存在，不再生成独立摘要。
    • 摘要层：若两条候选的摘要核心信息高度重合，必须重写或合并，确保每段摘要独立且有辨识度。
  • 不能依赖标题不同就视为不同条目；必须按事件事实和信息增量判断。

英文文章处理

  • 所有英文文章必须生成中文摘要。
  • 中文摘要必须准确覆盖原文核心信息、关键观点和重要结论，不能只翻译标题。
  • 语言要求简洁、逻辑清晰，不遗漏影响判断的关键信息。
  • 若原文是英文但正文无法完整访问，必须明确标记“英文原文未完整获取”，不能假装已完整总结。

链接格式

  • 所有链接统一使用 Markdown 格式：[标题文本](链接)
  • 标题文本必须简洁且能精准对应目标内容。
  • 禁止直接裸露 URL 作为最终输出主格式，除非是调试或原始抓取产物。

FreshRSS 配置约定

优先读取配置文件，其次读取环境变量。

推荐配置项：

  • base_url
  • username
  • api_password
  • output_json
  • output_md
  • limit
  • include_read

若既没有配置文件，也没有环境变量，不要假装已读取 FreshRSS，必须明确指出缺失项。

FreshRSS API 约定

  • API 地址：\x3Cbase_url>/api/greader.php
  • 登录接口：accounts/ClientLogin
  • 未读列表接口：reader/api/0/stream/contents/reading-list?xt=user/-/state/com.google/read&output=json
  • 已读标记接口：reader/api/0/edit-tag
  • 认证方式：Authorization: GoogleLogin auth=\x3CAuth>

默认优先输出两份原始产物：

  • digests/raw-freshrss.json
  • digests/raw-freshrss.md

然后默认继续输出“昨天窗口”切片：

  • digests/freshrss-yesterday.json
  • digests/freshrss-yesterday.md

再基于昨天窗口产物进入去重、聚类和摘要阶段。

已读回写状态

  • 当前默认流程不回写 FreshRSS 已读状态。
  • 生成 Daily Review、上传知识库、写入笔记后，也不要自动把源条目标记为已读。
  • 只有用户明确要求回写已读时，才允许进入该分支。
  • 若未来恢复该能力，也必须按“处理时间窗口”精确回写，不能清空整个原始未读池。
  • 需要人工触发时可使用 scripts/mark_freshrss_read.py，但默认流程禁止调用。

输出格式

每条候选至少包含：

  • 条目 ID
  • 标题
  • 来源
  • 时间
  • 链接（Markdown 格式）
  • 原始摘要或内容片段
  • 摘要
  • 主题标签
  • 聚类标记
  • 排序理由

排序原则

  • 公共重要性优先于个人偏好。
  • 个人相关性只作为弱信号。
  • 同一事件的多来源不应占满前排。

输出风格

  • 摘要尽量具体，避免空话。
  • 每段摘要必须与其他条目明显区分，禁止模板化重复表述。
  • 排序理由要可解释。
  • 若正文不可用，明确标记，不要装作已读过全文。

反例禁区

  • 只整理标题，不检查正文质量。
  • 摘要变成泛泛改写标题。
  • 多条摘要重复表达同一结论，却未合并或改写。
  • 用个人偏好压过公共重要性事件。
  • 还没做聚类就直接排序，导致同一事件反复出现。

参考文件

  • Digest 字段和过滤规则：references/digest-spec.md
  • FreshRSS Google Reader API 流程：references/freshrss-google-reader.md
  • FreshRSS 抓取脚本：scripts/fetch_freshrss_unread.py
  • FreshRSS 日期切片脚本：scripts/slice_freshrss_by_date.py
  • Digest 构建脚本：scripts/build_digest.py
  • FreshRSS 已读回写脚本：scripts/mark_freshrss_read.py
  • FreshRSS 配置样板：scripts/freshrss.config.example.json \r
Usage Guidance
This skill appears to do what it says (fetch FreshRSS unread items and build a deduplicated digest) but the registry failed to declare the FreshRSS credentials it needs. Before installing or running: (1) review the included Python scripts yourself or with a developer you trust; (2) expect to provide FRESHRSS_BASE_URL, FRESHRSS_USERNAME and FRESHRSS_API_PASSWORD (or a config file containing api_password) — do not supply your main login password if FreshRSS supports an API-specific password; (3) run the scripts in an isolated environment (container/VM) since they will make outbound HTTP requests to your FreshRSS instance and to article URLs to fetch bodies; (4) note that a mark-as-read helper exists but is not run by default — only run it if you explicitly want the skill to change your FreshRSS state; (5) if you need the registry to enforce least privilege, ask the publisher to update metadata to declare the required env vars and describe exactly what network calls the skill makes.
Capability Analysis
Type: OpenClaw Skill Name: digest-builder Version: 1.0.0 The skill bundle is designed to automate the collection and processing of RSS feeds via the FreshRSS API, which involves high-risk behaviors such as handling sensitive credentials (api_password) and performing automated web scraping of arbitrary URLs. Specifically, 'scripts/fetch_freshrss_unread.py' manages authentication tokens, and 'scripts/build_digest.py' uses urllib to fetch and parse content from external links found in the feed, which presents a significant SSRF (Server-Side Request Forgery) risk if the agent operates within a sensitive network. While these actions align with the stated purpose of building a structured digest, the combination of credential handling and unconstrained network fetching warrants a suspicious classification despite the lack of clear malicious intent.
Capability Assessment
✓ Purpose & Capability
Name, description, SKILL.md and included scripts consistently implement: fetching FreshRSS unread via the Google Reader API, URL deduplication, clustering, body fetching, noise filtering, summary generation and output to local files. The requested functionality matches the stated purpose.
✓ Instruction Scope
SKILL.md clearly restricts scope to FreshRSS fetch → slice → dedupe → summarize and explicitly says not to auto-mark items read. Instructions reference only FreshRSS endpoints, local files, and typical processing steps; they do not ask to read unrelated system files or to transmit data to unknown external endpoints.
✓ Install Mechanism
There is no install spec and all code is plain Python scripts included in the bundle. No external download/installation from untrusted URLs is present. This is the lowest-risk install pattern for a code-containing skill.
⚠ Credentials
The registry metadata declares no required environment variables or primary credential, but the SKILL.md and scripts explicitly expect FreshRSS configuration and an API password (FRESHRSS_BASE_URL, FRESHRSS_USERNAME, FRESHRSS_API_PASSWORD or a config JSON including api_password). That mismatch is problematic: the skill will fail without these secrets, and the metadata does not warn users or request them up front.
✓ Persistence & Privilege
The skill does not request always:true, does not persistently modify other skills or system settings, and by default does not mark items as read. A separate mark_freshrss_read script exists but only runs if explicitly invoked; the SKILL.md forbids automatic back-writing of read state by default.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install digest-builder
  3. After installation, invoke the skill by name or use /digest-builder
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of digest-builder. - Aggregates FreshRSS unread feed items via Google Reader API and processes them through de-duplication, event clustering, noise filtering, content validation, summary generation, and preliminary sorting. - Ensures strict deduplication across URL, event, and summary levels, focusing on information gain and source reliability. - Outputs results in both JSON and Markdown formats, with summaries and required metadata such as tags and sorting rationale. - Designed for structured, clean candidate pool creation—intended as a preprocessing step, not final daily digests. - Enforces clear handling of incomplete or inaccessible content and emphasizes transparent output formatting and configuration checks.
Metadata
Slug digest-builder
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Digest Builder?

将聚合池中的原始候选内容转换为可判断的 Digest：优先从 FreshRSS 的未读列表经 Google Reader API 拉取候选，再执行 URL 去重、相似事件聚类、正文抓取检查、噪音过滤、摘要生成与初步排序。用于需要把一批原始 feed 条目或 FreshRSS 未读条目整理成结构化候选文档的场景。 It is an AI Agent Skill for Claude Code / OpenClaw, with 94 downloads so far.

How do I install Digest Builder?

Run "/install digest-builder" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Digest Builder free?

Yes, Digest Builder is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Digest Builder support?

Digest Builder is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Digest Builder?

It is built and maintained by 阿晨聊技术 (@gcdd1993); the current version is v1.0.0.

More Skills
self-improving agent
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⬇ 463283 · by pskoett
Skill Vetter
Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
⬇ 259410 · by spclaudehome
Polymarket
Query Polymarket prediction markets. Check odds, find trending markets, search events, track price movements.
⬇ 232503 · by joelchance
Self-Improving + Proactive Agent
Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...
⬇ 200423 · by ivangdavila
Github
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
⬇ 191113 · by steipete
ontology
Typed knowledge graph for structured agent memory and composable skills. Use when creating/querying entities (Person, Project, Task, Event, Document), linkin...
⬇ 190156 · by oswalpalash

💬 Comments