claude-code-analyzer
Pass
Audited by Gen Agent Trust Hub on Mar 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Executes local bash scripts
scripts/analyze.shandscripts/analyze-claude-md.shto analyze usage patterns and project metadata. - [DATA_EXFILTRATION]: Accesses files and directories related to the Claude Code environment, including
~/.claude/projects(containing interaction history) and~/.claude/settings.json, to evaluate configurations and usage frequency. - [EXTERNAL_DOWNLOADS]: Fetches current documentation and engineering guidelines from official Claude and Anthropic domains (
docs.claude.com,anthropic.com) to ensure configurations follow latest best practices. - [PROMPT_INJECTION]: The skill processes untrusted interaction logs from past sessions, creating a surface for indirect prompt injection.
- Ingestion points: History JSONL files in
~/.claude/projects. - Boundary markers: No specific delimiters or markers to ignore embedded instructions were identified in the analyzed file.
- Capability inventory: Executes bash scripts, modifies project configuration files, and performs network requests via
web_fetch. - Sanitization: There is no mention of sanitization or validation of the ingested history data.
Audit Metadata