NYC

claude-code-analyzer

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The skill instructs the agent to execute local shell scripts (bash scripts/analyze.sh and bash scripts/analyze-claude-md.sh). These scripts are referenced in the workflow but their contents are not provided, meaning they could perform arbitrary operations on the user's system.
  • DATA_EXPOSURE (MEDIUM): The skill accesses sensitive information including tool usage history (JSONL files) and modifies the global ~/.claude/settings.json file. While intended for optimization, this provides a pathway for exposing or altering a user's operational history and security configurations.
  • EXTERNAL_DOWNLOADS (LOW): Uses the web_fetch tool to retrieve content from docs.claude.com and anthropic.com. Although these are trusted domains, the skill uses the retrieved content to dynamically generate and overwrite configuration files without a verification step.
  • INDIRECT_PROMPT_INJECTION (LOW):
  • Ingestion points: Fetches remote documentation and reads local project files (e.g., package.json, cargo.toml).
  • Boundary markers: None. The skill does not provide instructions to ignore potentially malicious content found in project files or fetched documentation.
  • Capability inventory: Includes bash command execution, file writing/modification, and modification of global tool settings.
  • Sanitization: No evidence of sanitization or validation of data before it is used to create or update configurations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:20 PM