geo-optimizer

Pass

Audited by Gen Agent Trust Hub on Aug 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell script using the find command to locate its reference file (geo-techniques.md). This search traverses specific hidden directories associated with the agent's environment, such as ~/.claude/plugins and ~/.codex/skills. While the search is restricted to a specific filename, it constitutes execution of shell logic to interact with the file system.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection. It is designed to fetch and analyze external content from URLs and local files provided by the user (Ingestion point: SKILL.md Step 3). The instructions do not include boundary markers or specific directives to ignore instructions that might be embedded within that external content (Sanitization: Absent). This could allow a malicious website to influence the agent's behavior during the audit or optimization process.
  • [EXTERNAL_DOWNLOADS]: The skill references external tools and GitHub repositories for citation tracking and measurement, including github.com/AI2HU/gego and github.com/jonradoff/llmopt. These are documented as resources for the user to verify AI citations but involve directing the user to third-party codebases.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 15, 2026, 03:14 PM
Security Audit — agent-trust-hub — geo-optimizer