skills/benwaffle/skills/claude-stats/Gen Agent Trust Hub

claude-stats

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill aggregates data from Claude Code log files (~/.claude/projects/*.jsonl), which include environment-derived strings like project paths (cwd), tool names, and skill names.
  • Ingestion points: The build_stats.py script recursively reads all JSONL files within the user's Claude projects directory.
  • Boundary markers: No boundary markers or sanitization logic is present to distinguish between logged data and potential malicious content within the logs.
  • Capability inventory: The skill utilizes Bash to execute the Python aggregation script and the open command to display the resulting HTML report.
  • Sanitization: The generated HTML report renders log data using innerHTML. This creates a Cross-Site Scripting (XSS) vulnerability if an attacker can influence the logs, for example, by creating a project in a directory with a malicious name containing script tags.
  • [COMMAND_EXECUTION]: The skill runs a Python script to process local log files and invokes the system's open command to launch the interactive report in the default web browser.
  • [EXTERNAL_DOWNLOADS]: The generated HTML report fetches the Chart.js library and a date-fns adapter from the jsdelivr CDN to render interactive visualizations.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 06:25 AM
Security Audit — agent-trust-hub — claude-stats