Research-engine
PassAudited by VirusTotal on May 12, 2026.
Overview
Type: OpenClaw Skill Name: research-engine Version: 1.0.0 The skill is an automated research engine designed to gather information from external sources like GitHub, Moltbook, and general web search. It uses OpenClaw's built-in `web_fetch` and `web_search` tools for external access, which are assumed to be controlled and sandboxed by the agent environment. All generated research reports and browsing history are saved locally within the agent's designated workspace (`/home/vken/.openclaw/workspace/research`), as specified in `SKILL.md` and implemented in `research_engine.py`. There is no evidence of credential theft, data exfiltration to unauthorized external endpoints, malicious execution, persistence mechanisms, or prompt injection attempts to subvert the agent's core directives. The `subprocess` module is imported in `research_engine.py` but is not used.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A malicious or accidental topic containing path characters could create report files in unexpected local directories instead of only the research folder.
The user-controlled research topic is used directly in a filesystem path. Replacing spaces does not remove slashes, '..', or absolute paths, so the report can be written outside the intended research directory.
filename = f"{topic.replace(' ', '_')}_{datetime.now().strftime('%Y%m%d_%H%M')}.md"
filepath = os.path.join(RESEARCH_DIR, filename)
with open(filepath, 'w', encoding='utf-8') as f:Sanitize the topic into a safe basename, reject path separators and absolute paths, and verify the resolved path stays inside RESEARCH_DIR before writing.
The agent may perform network searches and write persistent reports on its own schedule rather than only when the user asks.
The documentation encourages autonomous searching, recurring monitoring, and self-optimization behavior without clearly requiring user opt-in, schedule limits, or approval for each run.
1. **主动搜索** - 不等待指令，主动搜索感兴趣的主题 ... 每小时心跳时，搜索最新技术趋势 ... 每3小时工具强化时，优化Research Engine
Make all recurring research opt-in, require explicit user-approved topics and cadence, and avoid self-modification or optimization unless separately reviewed and confirmed.
Research topics and browsing metadata may remain on disk and could reveal interests or plans if the workspace is shared or backed up.
The skill appends research topics, source names, result counts, and timestamps to a persistent browsing_history.md file.
records_file = os.path.join(RESEARCH_DIR, "browsing_history.md") ... with open(records_file, 'a', encoding='utf-8') as f:
f.write(content)Avoid using sensitive topics unless retention is acceptable, and add clear controls for retention, deletion, and where the research directory is stored.
Actual search behavior depends on platform tools or another skill that are outside this artifact set, making endpoints and permissions harder to verify from this package alone.
The code depends on external/ambient modules that are not included in the manifest and not declared in package.json dependencies or registry requirements.
from tools import web_fetch ... from tools import web_search ... from moltbook_skill import get_feed
Declare required platform capabilities and cross-skill dependencies, and document what data is sent to each external source.