Claw News

The name/description (news aggregator with AI search, RSS, and push delivery) aligns with the included scripts (RSS fetcher, summarizer, search engine, digest generator). However the registry metadata claims no required environment variables or primary credential, while SKILL.md and multiple code files clearly expect several API keys (KIMI_API_KEY, MINIMAX_API_KEY, ANTHROPIC_API_KEY) and push gateway credentials. The presence of multiple LLM provider clients is coherent with a multi-provider search aggregator, but the metadata omission is inconsistent and should be corrected/clarified.

Runtime instructions and code focus on fetching RSS feeds, running web-search via configured LLM APIs, summarizing and formatting digests, caching results, and optionally pushing digests to channels (Slack/OpenClaw gateway). The SKILL.md and scripts reference only local config, feed sources, caches, and configured push endpoints; they do not instruct reading arbitrary system files or hidden data. The cron examples schedule local script execution. Review of the remaining truncated orchestrator scripts (newsman.py, search_engine.py, result_aggregator.py) is recommended to confirm no unexpected operations, but nothing in provided files indicates clandestine data collection beyond configured sources and delivery channels.

No install spec in the registry (instruction-only install). The README instructs creating a venv and pip installing requirements.txt — a normal, low-risk install path. requirements.txt lists standard libraries (openai, anthropic, requests, beautifulsoup4). There are no downloads from arbitrary URLs or extract/install steps that would write unknown binaries to disk.

Although requiring API keys for multiple LLM/search providers is reasonable for a multi-backend news search/summarization tool, the package metadata advertises no required env vars while SKILL.md and code require several sensitive values (KIMI_API_KEY, MINIMAX_API_KEY, ANTHROPIC_API_KEY, and GATEWAY_URL/GATEWAY_TOKEN for push). The gateway token in particular could allow posting messages through an OpenClaw gateway or similar — ensure tokens given are scoped and limited. The number and sensitivity of credentials should be documented in the registry metadata so users can decide before installing.

The skill does not request 'always: true' and is user-invocable only; it does provide cron examples for scheduled runs, which require explicit user action to set up. There is no evidence it attempts to modify other skills or system-wide agent settings. Autonomous invocation is allowed by platform defaults but is not combined here with other high-risk indicators.