Exa (1)

**Quick judgement** Useful, focused skill for neural web search and URL content extraction built on the Exa API. Strong for developers or power users who want scriptable, filterable web search (people, companies, news, research, code) from within the OpenClaw environment, but it’s tightly coupled to Exa, requires API setup, and provides no built-in safety or abuse controls. **What it does well** - Wraps Exa AI’s neural search in straightforward bash scripts (`search.sh`, `content.sh`) with environment-variable options (NUM, TYPE, CATEGORY, DOMAINS, EXCLUDE, SINCE, UNTIL, LOCATION). - Supports multiple content categories: people/LinkedIn-style, companies, news, research papers, GitHub, tweets, PDFs, financial reports. - Allows deep / comprehensive search (`TYPE=deep`) plus date and domain filters, which is valuable for time-bounded or source-bounded research. - Includes a separate content extraction command to fetch full page text for downstream processing. **Key risks & limitations** - **Privacy & people search**: The ability to search for people (e.g., LinkedIn-style queries) can be used for invasive profiling or doxxing if misused; there are no explicit safeguards, rate limits, or ethical usage guidance in the skill. - **Content risks**: As a general web search tool, it may surface harmful, sensitive, or disallowed content (e.g., extremism, adult content, medical misinformation) with no built-in filtering or classification. Any downstream system must enforce its own safety policies. - **Third-party dependency**: Functionality, performance, and reliability all depend on Exa’s API availability, pricing, quotas, and terms of service. If Exa changes its API or rate limits, the skill may degrade or break. - **Setup & environment assumptions**: Requires manual creation of a credentials file at `~/.clawdbot/credentials/exa/config.json` and a Unix-like environment with bash. This is fine for developers but non-trivial for non-technical users or Windows-only setups. **Recommended scenarios** - As a **backend search component** inside a larger research or summarization workflow (e.g., call Exa via this skill, then separately summarize or analyze the results). - For **competitive intelligence and market research**, where constrained-domain and date-filtered news/company searches are important. - For **technical and academic research**, combining `CATEGORY="research paper"` with domain filters and `TYPE=deep` to retrieve relevant literature and then extracting full text. - For **developer workflows**, where bash scripting and env-var configuration fit well into existing automation or pipelines. **Less suitable when** - You need strong built-in content safety, compliance checks, or moderation; these must be handled externally. - You require a GUI or highly user-friendly experience for non-technical users. - You need offline search or independence from third-party APIs; this skill is fully dependent on Exa.