GEO Tracker
Security checks across static analysis, malware telemetry, and agentic risk
Overview
GEO Tracker appears purpose-aligned, but it uses your AI-provider API keys, sends audit prompts to external services, and has some setup/coverage details users should verify.
Before installing, use a virtual environment with pinned provider SDK versions, supply separate API keys with usage limits, keep prompt libraries non-sensitive, verify which engines are actually supported, and review any OpenClaw cron schedule so it does not run or spend API credits unexpectedly.
Static analysis
No static analysis findings were reported for this release.
VirusTotal
66/66 vendors flagged this skill as clean.
Risk analysis
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 user could overestimate what engines were checked and make marketing decisions from incomplete coverage.
The implemented engine map covers four providers, while SKILL.md also advertises Google AI Overview and uses an `--engines all` example. Users should verify actual coverage before trusting reports as all-engine audits.
ENGINE_MAP = {
"chatgpt": query_chatgpt,
"perplexity": query_perplexity,
"gemini": query_gemini,
"claude": query_claude,
}Confirm the engine list used in each run, avoid relying on `--engines all` unless implemented, and treat Google AI Overview coverage as unverified from these scripts.
Supplying these keys lets the scripts make requests that may be billed to your provider accounts.
The skill asks users to provide API credentials for multiple AI providers. This is expected for its purpose, but it authorizes calls under the user’s accounts and is not declared in the registry metadata.
export OPENAI_API_KEY="sk-..." export PERPLEXITY_API_KEY="..." export GOOGLE_API_KEY="..." export ANTHROPIC_API_KEY="sk-ant-..."
Use separate, least-privileged API keys where possible, set provider spending limits, and avoid placing keys in shared shell history or logs.
Any confidential text placed in prompts may be transmitted to external AI providers.
Prompt-library lines are sent to external AI engines during audits. This is central to the skill, but prompt content crosses third-party provider data boundaries.
# These are the queries sent to AI engines during audits
Keep prompt files limited to non-sensitive marketing/search queries and review each provider’s data-retention and privacy terms.
Future package versions could change behavior or introduce dependency risk.
The manual install step uses unpinned third-party Python packages. These SDKs are purpose-aligned, but versions and provenance are not locked by the artifacts.
pip3 install openai anthropic google-generativeai
Install in a virtual environment, pin package versions, and use trusted package sources.
Scheduled runs may continue using API credits and generating reports until the schedule is changed or removed.
The skill suggests user-configured recurring audits. This is disclosed and purpose-aligned, but it creates ongoing automated provider calls.
Use OpenClaw cron to run daily/weekly audits: ``` Schedule a daily GEO audit for EZsite.ai at 9am ``` The agent will run the audit and report findings.
Only enable scheduling after confirming prompts, engines, and expected cost, and periodically review or disable the cron job.