best-practices-researcher

Pass

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands (e.g., bundle show) for repository exploration when native file tools are insufficient.
  • [EXTERNAL_DOWNLOADS]: The skill is designed to perform online research using web search and external tools (Context7 MCP) to fetch documentation and community standards.
  • [DATA_EXPOSURE]: The instructions direct the agent to search for and read SKILL.md files in sensitive locations, including user home directories (~/.claude/skills/, ~/.codex/skills/), which may contain private implementation details or project-specific logic.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes instructions from potentially untrusted sources (third-party skill files and external websites) and incorporates them into its own behavioral logic.
  • Ingestion points: Reads SKILL.md files from project subdirectories (.claude/skills/, .agents/skills/) and user home directories (~/.claude/skills/), and fetches content from external documentation sites.
  • Boundary markers: Absent. The instructions do not explicitly tell the agent to ignore or delimit instructions found within the processed files.
  • Capability inventory: Has access to shell execution, file reading, and file searching tools.
  • Sanitization: Absent. There is no evidence of validation or filtering of the content extracted from external skills before it is synthesized into recommendations.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 4, 2026, 01:30 AM