← 返回 Skills 市场

Favorites Curator

Name: Favorites Curator
Author: luw2007

作者 luw2007 · GitHub ↗ · v0.1.0 · MIT-0

cross-platform ✓ 安全检测通过

222

总下载

当前安装

版本数

在 OpenClaw 中安装

/install favorites-curator

功能描述

Build and maintain a local favorites catalog from installed repositories, apps, skills, extensions, and hooks. Use when the user asks to create, refresh, cur...

使用说明 (SKILL.md)

Favorites Curator

Maintain a file-backed catalog of useful local resources.

Do First

Run scripts/scan_favorites.py to refresh entries and the latest snapshot.
Run scripts/daily_digest.py to compare the latest snapshot with the previous one and produce a digest.
Run scripts/notify_candidates.py when the user wants short notification copy for notable items.

Paths

Entries: favorites/entries/
Snapshots: favorites/snapshots/
Reports: favorites/reports/
Cache: favorites/enrichment-cache.json
Skill code: skills/favorites-curator/
Naming and schema notes: references/naming-and-model.md

Covered Sources

~/ai git repositories
brew installed formulae and casks
/Applications and ~/Applications
~/.codex/skills
~/.claude/skills
~/.agents/skills
~/.openclaw/workspace/skills
~/.openclaw/extensions
~/.openclaw/hooks

Expected Workflow

Refresh catalog

python3 skills/favorites-curator/scripts/scan_favorites.py

Use --limit-source to iterate on one source:

python3 skills/favorites-curator/scripts/scan_favorites.py --limit-source brew

Generate daily digest

python3 skills/favorites-curator/scripts/daily_digest.py

This produces a Markdown report with:

new entries
updated entries
likely similar items worth deduping or cross-referencing
notable items worth surfacing

Emit notification copy

python3 skills/favorites-curator/scripts/notify_candidates.py

Use --top N to cap the number of lines.

Data Rules

Prefer local metadata over online enrichment.
Leave missing fields empty or mark them as inferred in field_sources.
Keep filenames ASCII, sortable, and machine-friendly.
Store one entry per file.
Update the same entry file on re-scan instead of creating duplicates.

Output Contract

Each entry file must keep these fields in frontmatter:

name
author
data_type
install_path
install_time
update_time
summary
simple_usage
source_url

Other helper fields are allowed when useful for automation.

Notes

The scripts are local-first and safe to run repeatedly.
brew info --json=v2 --installed is used once per scan; avoid tight cron loops.
Read references/naming-and-model.md before changing filename rules or schema.

安全使用建议

This skill is internally consistent with its stated goal: it scans local directories, writes a local catalog under favorites/, and optionally fetches metadata from GitHub or vendor pages. Before installing or running it, consider: 1) it will read many local paths (~/ai repos, various skills folders, /Applications) — avoid running it in workspaces that contain sensitive files you don't want examined; 2) it performs outbound HTTP(S) requests to GitHub and other source URLs for enrichment, which can reveal that your agent looked up those pages; if you want to avoid that, run the scripts offline or block network access for the process; 3) snapshots and reports are stored under favorites/ — inspect those files to ensure no sensitive content is being recorded; 4) if you need higher assurance, review the remainder of scan_favorites.py (the enrichment and URL handling code) and run the scripts in an isolated environment first. I have medium confidence because the code we saw matches its purpose and contains expected network enrichment behavior; confidence would rise if the truncated portion of scan_favorites.py were available for full review.

功能分析

Type: OpenClaw Skill Name: favorites-curator Version: 0.1.0 The favorites-curator bundle is a system inventory tool designed to catalog local applications, git repositories, and AI skills. It uses standard system commands (brew, git) and file system traversal to gather metadata, which it then enriches by fetching descriptions from the GitHub API or vendor homepages. While it possesses broad read access to local software directories and performs network requests, these actions are strictly aligned with its stated purpose of maintaining a local resource catalog. No evidence of credential theft, malicious execution, or prompt injection was found in files like scan_favorites.py or SKILL.md.

能力评估

✓ Purpose & Capability

Name/description match the code and instructions: the scripts scan local paths (git repos under ~/ai, brew, /Applications, several skills folders), write entries/snapshots/reports under favorites/, and provide digest/notification utilities. Network enrichment (GitHub API and vendor pages) is a reasonable optional capability for improving metadata and is documented in the references file.

ℹ Instruction Scope

SKILL.md directs the agent to run the included scripts which read many local directories and files (skill directories, app locations, brew info, repo metadata). The scanner also performs outbound HTTP(S) calls to GitHub and vendor homepages for optional enrichment. It does not require or send credentials, but fetching remote URLs can create outbound network traffic that may reveal which items exist on your machine to those remote hosts (typical for online enrichment).

✓ Install Mechanism

There is no install spec and no downloads; this is an instruction-plus-local-scripts skill. All code is bundled with the skill and nothing in the manifest attempts to fetch or install external code during install.

✓ Credentials

The skill declares no required env vars or credentials. The scripts optionally respect OPENCLAW_WORKSPACE to locate a workspace root. No other environment variables or secrets are requested. Network access is unauthenticated (public GitHub API calls), which is proportionate to optional enrichment.

✓ Persistence & Privilege

always is false and the skill does not request persistent elevated platform privileges. It writes data only under the workspace favorites/ tree and its own skill directory. It does not modify other skills or global agent configuration.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install favorites-curator
安装完成后，直接呼叫该 Skill 的名称或使用 /favorites-curator 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v0.1.0

Initial publish

元数据

Slug favorites-curator

版本 0.1.0

许可证 MIT-0

累计安装 1

当前安装数 1

历史版本数 1

常见问题