Agent Economy ReportAgent Economy ReportCAUTION version 2.0.0 · audited 2026-09-17
This package is a markdown skill that teaches an agent how to use the Exa MCP web search service. It does not contain executable code beyond an example shell command that registers an MCP server. The skill directs search and research requests to Exa's remote server and can pass an Exa API key as part of the MCP URL. The files themselves warn about API-key exposure and people/profile search privacy concerns. No credential exfiltration, hidden payloads, or persistence mechanisms were found. Because it registers an external service and documents putting an API key in a URL, users should review the setup before installing.
| Declared purpose | Neural web search, content extraction, company and people research, code search, and deep research via the Exa MCP server. |
| Observed behavior | The package provides documentation and examples for configuring the Exa MCP server, selecting search/research tools, and constructing queries. It includes a bash example to run 'openclaw mcp add exa --url ...' and optional API-key URLs. No hidden code or data collection is present. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| network | needed | The skill's entire purpose is network search and content retrieval through the remote Exa MCP server at https://mcp.exa.ai/mcp (SKILL.md line 10). Examples include crawling arxiv.org (SKILL.md line 254). |
| shell | needed | The setup section gives a bash example to register the external MCP server using the openclaw CLI: 'openclaw mcp add exa --url "https://mcp.exa.ai/mcp"' (SKILL.md line 22). |
| credentials | needed | Optional higher limits and all tools use an Exa API key appended to the MCP URL as 'exaApiKey=YOUR_EXA_KEY' (SKILL.md line 38). The skill card warns this can expose keys in URLs, configs, logs, or screenshots (skill-card.md line 25). |
| other | needed | The skill is declarative markdown guidance for tool selection, query patterns, and token efficiency; it does not include a local code implementation. |
| severity | finding |
|---|---|
| info | Declared purpose The skill front-matter declares web search, content extraction, people/company research, code search, and deep research via Exa MCP. SKILL.md:3 Neural web search, content extraction, company and people research, code search, and deep research via the Exa MCP server. |
| info | External MCP endpoint All agent activity is directed to Exa's remote MCP server; this is the core network dependency. SKILL.md:10 **MCP server:** `https://mcp.exa.ai/mcp` |
| low | Shell configuration command The setup section provides a bash example that registers the external MCP server using the openclaw CLI. SKILL.md:22 openclaw mcp add exa --url "https://mcp.exa.ai/mcp" |
| medium | API key placed in URL To unlock higher limits, the guide instructs appending the Exa API key as a URL query parameter, which can leak in logs or shared configs. SKILL.md:38 https://mcp.exa.ai/mcp?exaApiKey=YOUR_EXA_KEY |
| low | Optional tool activation URL Shows how to enable optional tools, including people search and crawling; this expands remote data collection capabilities. SKILL.md:43 https://mcp.exa.ai/mcp?exaApiKey=YOUR_KEY&tools=web_search_exa,web_search_advanced_exa,people_search_exa,crawling_exa,company_research_exa,get_code_context_exa, |
| medium | Credential-exposure warning The package itself acknowledges the main credential risk created by putting the API key in the MCP URL. skill-card.md:25 Risk: API keys may be exposed if embedded directly in shared MCP URLs, configuration files, logs, or screenshots. |
| medium | Self-reported suspicious security verdict The skill card states the release security verdict is suspicious and asks for review; this supports a cautious rather than safe disposition. skill-card.md:33 Risk: The release security verdict is suspicious and requires review before installation. |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| arxiv.org | not in the service index | SKILL.md:254 |
| clawhub.ai | not in the service index | skill-card.md:9 |
| dashboard.exa.ai | not in the service index | SKILL.md:12 |
| exa.ai | not in the service index | SKILL.md:13 |
| github.com | not in the service index | SKILL.md:14 |
| linkedin.com | not in the service index | SKILL.md:178 |
| mcp.exa.ai | not in the service index | SKILL.md:10 |
| Binaries invoked | bash |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 0 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 357 | 24f08ac90b30a098… |
_meta.json | 6 | 0d5437f86f15cf14… |
skill-card.md | 61 | 9159b3cc5d93b11e… |
For agents
JSON: https://agenteconomy.report/k/web-search-exa.audit.json · badge: https://agenteconomy.report/k/web-search-exa.audited.svg ·
skill rating: /k/web-search-exa · commission an audit of any skill: US$ 29 per version.
The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).