Multi Source Tech News Digest

Aggregates and scores technology news daily from 100+ RSS feeds, GitHub releases, and web sources, providing customizable, filtered tech news digests.

This skill mostly does what its name promises (aggregate tech news) but has several red flags you should consider before installing: - The code uses subprocess.run with dynamically constructed Python -c strings that embed config values (RSS URLs and repo strings). If those config values are modified by an attacker or come from untrusted input, they could lead to arbitrary code execution. Ask the author to remove subprocess -c usage and call requests/feedparser directly in-process. - Some advertised sources (Twitter/X, web scraping) are mentioned in descriptions/config but are not actually implemented — verify the feature set if you need those sources. - The GitHub endpoints in config appear incorrect/unexpected; confirm how GitHub data is fetched and whether authentication is required for your use case. - test_skill.py contains a hardcoded absolute path (/home/pan/...) — likely a leftover from development. That is not a direct runtime threat but indicates the repo wasn't fully cleaned. - Because the skill makes outbound HTTP requests, run it in a sandboxed environment or restricted network if you want to limit data exfiltration risk, and review/replace the subprocess calls before granting production use. If you decide to proceed: require the maintainer to (1) replace subprocess -c calls with in-process library calls that properly validate/escape inputs, (2) fix/clarify GitHub source usage, (3) remove development hardcoded paths, and (4) document exactly which external endpoints will be contacted and whether credentials are ever required.

Type: OpenClaw Skill Name: multi-source-news-digest Version: 1.0.0 The `skill.py` file contains a critical code injection vulnerability. It uses `subprocess.run` to execute Python code, directly embedding user-configurable URLs (`rss_sources` and `github_repos` from `config.json`) into f-strings. As `SKILL.md` explicitly states that `config.json` is user-editable, an attacker could modify these URLs to inject arbitrary Python code, leading to Remote Code Execution (RCE) on the agent. While the provided `config.json` contains benign URLs, the underlying implementation flaw makes the skill highly risky.